From: levy
Subject: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1193939265.339517.43140@d55g2000hsg.googlegroups.com>
We are a small group of friends (3-5) who left the "professional"
programming scene about one and a half year ago to start our own Lisp
business.

In July, 2007 we started working on a project for the Hungarian
government with a very tight deadline. In mid October the system
showed first life signs and now we can say it is extensively used by
thousands of users. In spite of the short timeframe and the remaining
glitches it was a surprise even for us that the system is able to
handle such a load.

Unfortunately the application is not publicly accessible, so we can't
provide a URL, but we can give some of the details. It is used for
gathering data from the Hungarian communes to help the budget
planning.

The entire system is based on opensource software, it runs on multiple
SBCLs behind an apache load balancer. The data is stored in a
PostgreSQL server.

We have around 4000 registered users, average 300 online and more than
500 at peak times. There are 7000-10000 logins per day.

We use the following Common Lisp libraries. Some of which are ours and
some of which are built by others. Many thanks for their efforts,
especially to the SBCL developers!

  - Data storage/persistence
    - cl-perec
    - cl-rdbms
    - cl-postgres (postmodern)

  - Web GUI
    - cl-dwim (the umbrella project)
    - UCW
    - yaclml
    - parenscript
    - rfc*

  - PDF exporting
    - cl-pdf
    - cl-typesetting

  - Unit testing
    - stefil
    - SLIME and emacs with customizations

  - MOP
    - computed-class
    - contextl

  - Misc
    - bind
    - alexandria
    - iterate
    - cl-def
    - cl-l10n
    - cl-graphviz
    - cl-serializer
    - defclass-star
    - local-time
    - cffi
    - babel
    - verrazano

and more...

Main software components:
  - Ubuntu linux 6.10
  - PostgreSQL 8.2
  - SBCL 1.0.10
  - Apache2 (due to various headaches soon to be replaced by a lisp
based load balancer)

Hardware:
  - 10 nodes with two dual core x86-64 CPUs summing up to 40
cores altogether, each node has 3 Gbytes ram.
  - Currently we use one of the nodes as database server but soon it
will
be replaced with another server with 2 or 4 quad core x86-64 CPUs, 300
GByte disk in raid 5 and 16 GBytes ram.

Some people say Lisp is dead but seems like there's life after
death... ;-)

From: Ken Tilton
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <BAsWi.46$6I6.40@newsfe08.lga>
levy wrote:
> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
> 
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
> 
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.
> 
> The entire system is based on opensource software, it runs on multiple
> SBCLs behind an apache load balancer. The data is stored in a
> PostgreSQL server.
> 
> We have around 4000 registered users, average 300 online and more than
> 500 at peak times. There are 7000-10000 logins per day.
> 
> We use the following Common Lisp libraries. Some of which are ours and
> some of which are built by others. Many thanks for their efforts,
> especially to the SBCL developers!
> 
>   - Data storage/persistence
>     - cl-perec
>     - cl-rdbms
>     - cl-postgres (postmodern)
> 
>   - Web GUI
>     - cl-dwim (the umbrella project)
>     - UCW
>     - yaclml
>     - parenscript
>     - rfc*
> 
>   - PDF exporting
>     - cl-pdf
>     - cl-typesetting
> 
>   - Unit testing
>     - stefil
>     - SLIME and emacs with customizations
> 
>   - MOP
>     - computed-class
>     - contextl
> 
>   - Misc
>     - bind
>     - alexandria
>     - iterate
>     - cl-def
>     - cl-l10n
>     - cl-graphviz
>     - cl-serializer
>     - defclass-star
>     - local-time
>     - cffi
>     - babel
>     - verrazano
> 
> and more...
> 
> Main software components:
>   - Ubuntu linux 6.10
>   - PostgreSQL 8.2
>   - SBCL 1.0.10
>   - Apache2 (due to various headaches soon to be replaced by a lisp
> based load balancer)
> 
> Hardware:
>   - 10 nodes with two dual core x86-64 CPUs summing up to 40
> cores altogether, each node has 3 Gbytes ram.
>   - Currently we use one of the nodes as database server but soon it
> will
> be replaced with another server with 2 or 4 quad core x86-64 CPUs, 300
> GByte disk in raid 5 and 16 GBytes ram.
> 
> Some people say Lisp is dead but seems like there's life after
> death... ;-)
> 

And you owe it all to Cells, yes?

I think we want to hang onto the dead status, pick a Grateful Dead song 
as our theme, maybe translate the CLHS into Latin, resist at all cost 
becoming au courant, reopen the spec to make it less palatable if 
necessary. We don't want to end up like Britney Spears.

kzo

-- 
http://www.theoryyalgebra.com/

"In the morning, hear the Way;
  in the evening, die content!"
                     -- Confucius
From: Jason
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1193963434.509285.183080@v29g2000prd.googlegroups.com>
On Nov 1, 3:38 pm, Ken Tilton <···········@optonline.net> wrote:
> levy wrote:
> > We are a small group of friends (3-5) who left the "professional"
> > programming scene about one and a half year ago to start our own Lisp
> > business.
>
> > In July, 2007 we started working on a project for the Hungarian
> > government with a very tight deadline. In mid October the system
> > showed first life signs and now we can say it is extensively used by
> > thousands of users. In spite of the short timeframe and the remaining
> > glitches it was a surprise even for us that the system is able to
> > handle such a load.
>
> > Unfortunately the application is not publicly accessible, so we can't
> > provide a URL, but we can give some of the details. It is used for
> > gathering data from the Hungarian communes to help the budget
> > planning.
>
> > The entire system is based on opensource software, it runs on multiple
> > SBCLs behind an apache load balancer. The data is stored in a
> > PostgreSQL server.
>
> > We have around 4000 registered users, average 300 online and more than
> > 500 at peak times. There are 7000-10000 logins per day.
>
> > We use the following Common Lisp libraries. Some of which are ours and
> > some of which are built by others. Many thanks for their efforts,
> > especially to the SBCL developers!
>
> >   - Data storage/persistence
> >     - cl-perec
> >     - cl-rdbms
> >     - cl-postgres (postmodern)
>
> >   - Web GUI
> >     - cl-dwim (the umbrella project)
> >     - UCW
> >     - yaclml
> >     - parenscript
> >     - rfc*
>
> >   - PDF exporting
> >     - cl-pdf
> >     - cl-typesetting
>
> >   - Unit testing
> >     - stefil
> >     - SLIME and emacs with customizations
>
> >   - MOP
> >     - computed-class
> >     - contextl
>
> >   - Misc
> >     - bind
> >     - alexandria
> >     - iterate
> >     - cl-def
> >     - cl-l10n
> >     - cl-graphviz
> >     - cl-serializer
> >     - defclass-star
> >     - local-time
> >     - cffi
> >     - babel
> >     - verrazano
>
> > and more...
>
> > Main software components:
> >   - Ubuntu linux 6.10
> >   - PostgreSQL 8.2
> >   - SBCL 1.0.10
> >   - Apache2 (due to various headaches soon to be replaced by a lisp
> > based load balancer)
>
> > Hardware:
> >   - 10 nodes with two dual core x86-64 CPUs summing up to 40
> > cores altogether, each node has 3 Gbytes ram.
> >   - Currently we use one of the nodes as database server but soon it
> > will
> > be replaced with another server with 2 or 4 quad core x86-64 CPUs, 300
> > GByte disk in raid 5 and 16 GBytes ram.
>
> > Some people say Lisp is dead but seems like there's life after
> > death... ;-)
>
> And you owe it all to Cells, yes?
>
> I think we want to hang onto the dead status, pick a Grateful Dead song
> as our theme, maybe translate the CLHS into Latin, resist at all cost
> becoming au courant, reopen the spec to make it less palatable if
> necessary. We don't want to end up like Britney Spears.
>

My refusal to wear undergarments has no bearing upon my ability to
program, thank you very much!
From: Kamen TOMOV
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <upryt7s4p.fsf@cybuild.com>
On Thu, Nov 01 2007, levy wrote:

> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own
> Lisp business.
>
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the
> remaining glitches it was a surprise even for us that the system is
> able to handle such a load.

The high load ability will not be a news for us. The real news is that
you won the contract over gians in the industry. Congratulations! How
did you do that? Please tell us what should we do to get a contract
with our governments?

Bulgaria used to be the leader in the IT industry of the so called
"Soviet camp" and we are followers now - all the serious work here is
done in Java and .NET.

-- 
�����
From: Dimiter "malkia" Stanev
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <5ovi4aFoofphU1@mid.individual.net>
Kamen TOMOV wrote:
> On Thu, Nov 01 2007, levy wrote:
> 
>> We are a small group of friends (3-5) who left the "professional"
>> programming scene about one and a half year ago to start our own
>> Lisp business.
>>
>> In July, 2007 we started working on a project for the Hungarian
>> government with a very tight deadline. In mid October the system
>> showed first life signs and now we can say it is extensively used by
>> thousands of users. In spite of the short timeframe and the
>> remaining glitches it was a surprise even for us that the system is
>> able to handle such a load.
> 
> The high load ability will not be a news for us. The real news is that
> you won the contract over gians in the industry. Congratulations! How
> did you do that? Please tell us what should we do to get a contract
> with our governments?
> 
> Bulgaria used to be the leader in the IT industry of the so called
> "Soviet camp" and we are followers now - all the serious work here is
> done in Java and .NET.
> 

Back in the days, I remember dBase, Clipper, and then Turbo 
Pascal/Delphi to rule where I was in Bulgaria (at least that was the 
case in the city of Burgas - The Busoft/Prosist company).
From: Kamen TOMOV
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <ulk9hosh5.fsf@cybuild.com>
On Fri, Nov 02 2007, Dimiter \"malkia\" Stanev wrote:

> Kamen TOMOV wrote:
>> ...
>>
>> Bulgaria used to be the leader in the IT industry of the so called
>> "Soviet camp" and we are followers now - all the serious work here
>> is done in Java and .NET.
>>
>
> Back in the days, I remember dBase, Clipper, and then Turbo
> Pascal/Delphi to rule where I was in Bulgaria (at least that was the
> case in the city of Burgas - The Busoft/Prosist company).
>

Generally, we used to be a leader in computer manufacturing as we
supplied 40 percent of the computers in COMECON (���) [1] including
personal computers IMKO/Pravetz (Bulgarian clone of 6502) and
mainframes IZOT [2] (clones of IBM's System/360, System/370 and
System/390).

The electronics industry employed 300,000 workers, and it generated 8
billion rubles a year (US$13.3 billion) [3].

Beside Pascal and Clipper most of the development here used to be in
assembly language, C, Pascal and C++. I guess like everywhere though,
I would underline assembly language as related to the computer
manufacturing.

Perhaps some of the reasons why COMECON decided that we would be so
important IT center were the high quality education system and the
fact that John Atanassov (from Bulgarian origin) [5] invented the
first electronic digital computer. In other words he was the one who
decided to apply electronics in computers.

References

[1] http://en.wikipedia.org/wiki/Comecon
[2] http://en.wikipedia.org/wiki/ES_EVM
[3] http://en.wikipedia.org/wiki/History_of_computer_hardware_in_Soviet_Bloc_countries
[4] http://en.wikipedia.org/wiki/Computer
[5] http://en.wikipedia.org/wiki/John_Vincent_Atanasoff

-- 
�����
From: Slobodan Blazeski
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1193953028.233322.310830@50g2000hsm.googlegroups.com>
On Nov 1, 10:47 am, levy <················@gmail.com> wrote:
> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
>
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
>
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.
>
> The entire system is based on opensource software, it runs on multiple
> SBCLs behind an apache load balancer. The data is stored in a
> PostgreSQL server.
>
> We have around 4000 registered users, average 300 online and more than
> 500 at peak times. There are 7000-10000 logins per day.
>
> We use the following Common Lisp libraries. Some of which are ours and
> some of which are built by others. Many thanks for their efforts,
> especially to the SBCL developers!
>
>   - Data storage/persistence
>     - cl-perec
>     - cl-rdbms
>     - cl-postgres (postmodern)
>
>   - Web GUI
>     - cl-dwim (the umbrella project)
>     - UCW
>     - yaclml
>     - parenscript
>     - rfc*
>
>   - PDF exporting
>     - cl-pdf
>     - cl-typesetting
>
>   - Unit testing
>     - stefil
>     - SLIME and emacs with customizations
>
>   - MOP
>     - computed-class
>     - contextl
>
>   - Misc
>     - bind
>     - alexandria
>     - iterate
>     - cl-def
>     - cl-l10n
>     - cl-graphviz
>     - cl-serializer
>     - defclass-star
>     - local-time
>     - cffi
>     - babel
>     - verrazano
>
> and more...
>
> Main software components:
>   - Ubuntu linux 6.10
>   - PostgreSQL 8.2
>   - SBCL 1.0.10
>   - Apache2 (due to various headaches soon to be replaced by a lisp
> based load balancer)
>
> Hardware:
>   - 10 nodes with two dual core x86-64 CPUs summing up to 40
> cores altogether, each node has 3 Gbytes ram.
>   - Currently we use one of the nodes as database server but soon it
> will
> be replaced with another server with 2 or 4 quad core x86-64 CPUs, 300
> GByte disk in raid 5 and 16 GBytes ram.
>
> Some people say Lisp is dead but seems like there's life after
> death... ;-)

Congratulations , post some screenshots or video to see it in action.
cheers
Slobodan
From: Rainer Joswig
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <joswig-24916C.22413701112007@news-europe.giganews.com>
In article <·······················@d55g2000hsg.googlegroups.com>,
 levy <················@gmail.com> wrote:

> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
> 
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
> 
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.

...

> Hardware:
>   - 10 nodes with two dual core x86-64 CPUs summing up to 40
> cores altogether, each node has 3 Gbytes ram.
>   - Currently we use one of the nodes as database server but soon it
> will
> be replaced with another server with 2 or 4 quad core x86-64 CPUs, 300
> GByte disk in raid 5 and 16 GBytes ram.
> 
> Some people say Lisp is dead but seems like there's life after
> death... ;-)

Sounds impressive. Congratulations!

Why do you need ten nodes with dual core machines? Is there a
lot of computation going on? Are they all running the same
software or is there some structure to the deployed system
(web frontend, business logic, database, ...)?
From: levy
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1193990610.926181.99210@57g2000hsv.googlegroups.com>
> Why do you need ten nodes with dual core machines? Is there a
> lot of computation going on? Are they all running the same
> software or is there some structure to the deployed system
> (web frontend, business logic, database, ...)?

Currently 9 nodes are running the same Lisp image (BTW: the system
starts within 10 seconds) but they read the configuration from the
database which may be different in the future (not yet). The 10th node
which is going to be replaced soon is running apache and postgres
only.

By high load we meant several concurrent sessions (50 per node)
without big troubles on the Lisp side. At the moment hardware is not
fully utilized but the project is going to expand next year so we were
planning ahead. We also do some parallel computations which are stored
as continuations in the database and spread across the 40 cores.
Maintenance and flexibility are more important for us than being super
optimized and hardware (the total cost so far is $15,000) is sometimes
cheaper than software.

As for the contract I think we were pretty lucky to win, so we can't
give you the silver bullet other than being cheap and flexible beyond
limits and have somebody really motivated on the other side. ;-)
From: bob
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194166566.754251.146750@k35g2000prh.googlegroups.com>
On Nov 2, 1:03 am, levy <················@gmail.com> wrote:
> > Why do you need ten nodes with dual core machines? Is there a
> > lot of computation going on? Are they all running the same
> > software or is there some structure to the deployed system
> > (web frontend, business logic, database, ...)?
>
> Currently 9 nodes are running the same Lisp image (BTW: the system
> starts within 10 seconds) but they read the configuration from the
> database which may be different in the future (not yet). The 10th node
> which is going to be replaced soon is running apache and postgres
> only.
>
> By high load we meant several concurrent sessions (50 per node)
> without big troubles on the Lisp side. At the moment hardware is not
> fully utilized but the project is going to expand next year so we were
> planning ahead. We also do some parallel computations which are stored
> as continuations in the database and spread across the 40 cores.
> Maintenance and flexibility are more important for us than being super
> optimized and hardware (the total cost so far is $15,000) is sometimes
> cheaper than software.
>
> As for the contract I think we were pretty lucky to win, so we can't
> give you the silver bullet other than being cheap and flexible beyond
> limits and have somebody really motivated on the other side. ;-)

I was wondering if you have any figures on the limits these servers
can handle, and whether the bottleneck is in the CPU or memory
(assuming you are storing the sessions in memory?).

Thanks

Bob
From: levy
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194268519.636530.212810@19g2000hsx.googlegroups.com>
> I was wondering if you have any figures on the limits these servers
> can handle, and whether the bottleneck is in the CPU or memory
> (assuming you are storing the sessions in memory?).
>
> Thanks
>
> Bob
We don't have such figures at the moment. What I can say now is that
memory seems to be a bigger issue for us than CPU limits but we
haven't spent too much time on optimization yet. We have some minor gc/
application issues which are causing objects to stay in memory, they
are still to be solved. Automatic restarts at night seems to address
this for now.

levy
From: Alex Mizrahi
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <472b509f$0$90265$14726298@news.sunsite.dk>
 l> Hardware:
 l>   - 10 nodes with two dual core x86-64 CPUs summing up to 40
 l> cores altogether, each node has 3 Gbytes ram.

40 cores for 300 users online in a web application?
this sounds like a *terrible* waste of resources for me..

certainly it's hard to speak about absolute figures, but there is 
SPECweb2005, for example, and there single Xeon 3070 2.6 GHz dual core 8GB 
RAM has result of something like 10000 concurrent sessions.

but certainly it's cool to get system working within such tight deadline 
From: levy
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194023380.455619.114660@y42g2000hsy.googlegroups.com>
On nov. 2, 17:30, "Alex Mizrahi" <········@users.sourceforge.net>
wrote:
> 40 cores for 300 users online in a web application?
> this sounds like a *terrible* waste of resources for me..
>
> certainly it's hard to speak about absolute figures, but there is
> SPECweb2005, for example, and there single Xeon 3070 2.6 GHz dual core 8GB
> RAM has result of something like 10000 concurrent sessions.
10000 concurrent sessions in a non trivial web application might be
quite a lot, it strongly depends on what the application is doing.

Well, I'm not saying that our system could not be improved on that, in
fact we did not even bother to optimize things too much due to the
short timeframe. On the other hand if one calculates the relative
costs of hardware vs. software one might realize that it's not worth
spending too much time to be a few times faster.

levy

--
There's no perfectoin
From: ·············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194048946.615675.95340@v23g2000prn.googlegroups.com>
Wow! Pretty cool!

How do you handle UCW continuations when the user's session gets
redirected to a different box by the load balancer?   Are you making
sessions sticky to the back instance? Also, how do you handle hardware
failure -- what happens to continuations when the specific service
goes down?

Yarek


On Nov 1, 10:47 am, levy <················@gmail.com> wrote:
> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
>
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
>
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.
>
> The entire system is based on opensource software, it runs on multiple
> SBCLs behind an apache load balancer. The data is stored in a
> PostgreSQL server.
>
> We have around 4000 registered users, average 300 online and more than
> 500 at peak times. There are 7000-10000 logins per day.
>
> We use the following Common Lisp libraries. Some of which are ours and
> some of which are built by others. Many thanks for their efforts,
> especially to the SBCL developers!
>
>   - Data storage/persistence
>     - cl-perec
>     - cl-rdbms
>     - cl-postgres (postmodern)
>
>   - Web GUI
>     - cl-dwim (the umbrella project)
>     - UCW
>     - yaclml
>     - parenscript
>     - rfc*
>
>   - PDF exporting
>     - cl-pdf
>     - cl-typesetting
>
>   - Unit testing
>     - stefil
>     - SLIME and emacs with customizations
>
>   - MOP
>     - computed-class
>     - contextl
>
>   - Misc
>     - bind
>     - alexandria
>     - iterate
>     - cl-def
>     - cl-l10n
>     - cl-graphviz
>     - cl-serializer
>     - defclass-star
>     - local-time
>     - cffi
>     - babel
>     - verrazano
>
> and more...
>
> Main software components:
>   - Ubuntu linux 6.10
>   - PostgreSQL 8.2
>   - SBCL 1.0.10
>   - Apache2 (due to various headaches soon to be replaced by a lisp
> based load balancer)
>
> Hardware:
>   - 10 nodes with two dual core x86-64 CPUs summing up to 40
> cores altogether, each node has 3 Gbytes ram.
>   - Currently we use one of the nodes as database server but soon it
> will
> be replaced with another server with 2 or 4 quad core x86-64 CPUs, 300
> GByte disk in raid 5 and 16 GBytes ram.
>
> Some people say Lisp is dead but seems like there's life after
> death... ;-)
From: bob
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194053745.674120.249690@z24g2000prh.googlegroups.com>
Hi Levy,

"cl-serializer"

Is this what you used to serialize the continuations?  The only info I
could find is an empty mailing list.

Bob
From: ··············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194115181.572435.125090@22g2000hsm.googlegroups.com>
> "cl-serializer"
>
> Is this what you used to serialize the continuations?  The only info I
> could find is an empty mailing list.

http://common-lisp.net/cgi-bin/darcsweb/darcsweb.cgi?r=cl-serializer-cl-serializer;a=summary

that's all that is available online for now. check out the test file
for some info.
From: levy
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194268090.018484.151610@o80g2000hse.googlegroups.com>
On Nov 3, 2:35 am, bob <··········@gmail.com> wrote:
> Hi Levy,
>
> "cl-serializer"
>
> Is this what you used to serialize the continuations?  The only info I
> could find is an empty mailing list.
>
> Bob

Yes, this is yet another general purpose serializer library. For
others see cl-store or the ones you can find in elephant and rucksack.
The reason why we created a new one is that we needed better control
over funcionality and performance and there's no standalone lib other
than cl-store which is not that fast.

BTW, the actual continuations are created using arnesi and used to
load balance long running computations.

levy
From: ··············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194115508.484206.34540@50g2000hsm.googlegroups.com>
> How do you handle UCW continuations when the user's session gets
> redirected to a different box by the load balancer?   Are you making
> sessions sticky to the back instance? Also, how do you handle hardware

sure, sticky sessions.

> failure -- what happens to continuations when the specific service
> goes down?

if a node fails, the users logged in to that node will swear and we
ask for pardon. that's how life goes with non-mission-critical
apps... :)

- attila
From: ············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194057574.160859.107370@z9g2000hsf.googlegroups.com>
That's pretty cool. Congratulations.


On Nov 1, 12:47 pm, levy <················@gmail.com> wrote:
> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
>
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
>
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.
>
> The entire system is based on opensource software, it runs on multiple
> SBCLs behind an apache load balancer. The data is stored in a
> PostgreSQL server.
>
> We have around 4000 registered users, average 300 online and more than
> 500 at peak times. There are 7000-10000 logins per day.
>
> We use the following Common Lisp libraries. Some of which are ours and
> some of which are built by others. Many thanks for their efforts,
> especially to the SBCL developers!
>
>   - Data storage/persistence
>     - cl-perec
>     - cl-rdbms
>     - cl-postgres (postmodern)
>
>   - Web GUI
>     - cl-dwim (the umbrella project)
>     - UCW
>     - yaclml
>     - parenscript
>     - rfc*
>
>   - PDF exporting
>     - cl-pdf
>     - cl-typesetting
>
>   - Unit testing
>     - stefil
>     - SLIME and emacs with customizations
>
>   - MOP
>     - computed-class
>     - contextl
>
>   - Misc
>     - bind
>     - alexandria
>     - iterate
>     - cl-def
>     - cl-l10n
>     - cl-graphviz
>     - cl-serializer
>     - defclass-star
>     - local-time
>     - cffi
>     - babel
>     - verrazano
>
> and more...
>
> Main software components:
>   - Ubuntu linux 6.10
>   - PostgreSQL 8.2
>   - SBCL 1.0.10
>   - Apache2 (due to various headaches soon to be replaced by a lisp
> based load balancer)
>
> Hardware:
>   - 10 nodes with two dual core x86-64 CPUs summing up to 40
> cores altogether, each node has 3 Gbytes ram.
>   - Currently we use one of the nodes as database server but soon it
> will
> be replaced with another server with 2 or 4 quad core x86-64 CPUs, 300
> GByte disk in raid 5 and 16 GBytes ram.
>
> Some people say Lisp is dead but seems like there's life after
> death... ;-)
From: Pascal Costanza
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <5p3rs5Fph7guU1@mid.individual.net>
levy wrote:
> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
> 
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
> 
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.
> 
> The entire system is based on opensource software, it runs on multiple
> SBCLs behind an apache load balancer. The data is stored in a
> PostgreSQL server.
> 
> We have around 4000 registered users, average 300 online and more than
> 500 at peak times. There are 7000-10000 logins per day.
> 
> We use the following Common Lisp libraries. Some of which are ours and
> some of which are built by others. Many thanks for their efforts,
> especially to the SBCL developers!
> 
>   - Data storage/persistence
>     - cl-perec
>     - cl-rdbms
>     - cl-postgres (postmodern)
> 
>   - Web GUI
>     - cl-dwim (the umbrella project)
>     - UCW
>     - yaclml
>     - parenscript
>     - rfc*
> 
>   - PDF exporting
>     - cl-pdf
>     - cl-typesetting
> 
>   - Unit testing
>     - stefil
>     - SLIME and emacs with customizations
> 
>   - MOP
>     - computed-class
>     - contextl
> 
>   - Misc
>     - bind
>     - alexandria
>     - iterate
>     - cl-def
>     - cl-l10n
>     - cl-graphviz
>     - cl-serializer
>     - defclass-star
>     - local-time
>     - cffi
>     - babel
>     - verrazano
> 
> and more...

That's a lot of libraries that you use there. Did you ever have any 
serious integration problems, like name conflicts, etc.? Or did this all 
work smoothly?


Pascal

-- 
My website: http://p-cos.net
Common Lisp Document Repository: http://cdr.eurolisp.org
Closer to MOP & ContextL: http://common-lisp.net/project/closer/
From: ··············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194115080.609419.73280@d55g2000hsg.googlegroups.com>
> That's a lot of libraries that you use there. Did you ever have any
> serious integration problems, like name conflicts, etc.? Or did this all
> work smoothly?

many of them are our own libs and we have the commit bit for many of
the others. so integration wasn't such a big problem.

as you can see if things turned out to be a headache with a library we
were not hesitating to patch it or rewrite it according to our taste.
(cl-serializer and cl-rdbms are two good examples for reimplementing
an existing functionality)

but obviously this kind of thinking also resulted in clashes and forks
where the way we see stuff didn't match the author's view (luckily it
only happened with one project: slime).

but i think it was smoother then what people would think. to put it
mildly, some part of the lisp community is not such a big fan of many
dependencies. i think this is mostly due to the rough support for
dependency/version/install/update management for libs. but things get
manageable once you start using a distributed VCS and get used to the
rough edges of the dependency management.

- attila
From: Antony Sequeira
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <4OSdnT12wYI50azanZ2dnUVZ_rbinZ2d@comcast.com>
Pascal Costanza wrote:
> That's a lot of libraries that you use there. Did you ever have any 
> serious integration problems, like name conflicts, etc.? Or did this all 
> work smoothly?
> 
That is *not* a lot of libraries. For any real world app one does need
lots of stuff and uses libs wherever possible.
I have been looking to use lisp but it is this kind of comments that
make me nervous.
This is not meant to pick on you or your specific response. It is a
honest comment about how I feel. The purpose is to hopefully make people
realize that libs are as important as  is the core language
capabilities. I'd say the 'success' (for some measure) of java owes much
to libs, portability, scalability (in terms of processors and memory at
least) among other things, even though the core language is poor.

And for those who think documentation is not important, think about
these guys having to read the source code of all those libs where there
is no documentation.

My day job (java) i work on something that does low double digit web 
requests per second doing fairly compute intensive stuff and there are 
more than one box doing it so it is a few millions of requests per day.
So I also get nervous (assuming i am right) that sbcl is the only truly
multi-threaded lisp (having a big lock doesn't count)

I also get worried hearing that using a 1.5GB heap causes issues, while
we routinely use way more than that in Java. I do not want to be
specific with numbers, cause I don't want trouble at my work.

There are algorithmic complexities and then there are people+engineering
complexities.  I get the impression that lots of people here have little
clue to the people+engineering complexities while they can run circles
round me about say NLP or continuations for example.

Also things like 80% of time is spent in 20% of code may not be a very 
helpful rule of thumb when a web request needs to to completed in less 
than a  second and you want to be able to run a few of those 
simultaneously. What is acceptable for a desktop app is not necessarily 
acceptable in a web app.

I know this not all 100% coherent and logical, but this usenet and I am 
not provding answer to a technical question :)

On a positive note, I am much more confident about the availability of
libs for CL today than say 2 years ago.

I am going to be asking (and have already asked and gotten no reply
regarding local-time) a few technical questions here as time goes and
hopefully this does not get me black listed. I do value and appreciate
your contributions as well as others.

-Antony
From: ··············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194427062.913513.41500@y42g2000hsy.googlegroups.com>
> That is *not* a lot of libraries. For any real world app one does need


that was only the list of the important ones, the total is about 60.


> capabilities. I'd say the 'success' (for some measure) of java owes much
> to libs, portability, scalability (in terms of processors and memory at
> least) among other things, even though the core language is poor.


i agree in that and that's why i'm sorry that the infrastructure
around library handling is not too great in lisp.


> And for those who think documentation is not important, think about
> these guys having to read the source code of all those libs where there
> is no documentation.


heh, all of us are one of those guys! :) i _almost never_ read docs!
exception is when i need to get a bird's view of what the project is
all about, but those paragraphs i'm looking for then are not those
that you mean by docs. lisp can be written so that it's better
readable than a handwritten documentation, and the code itself is
never outdated!

if you ever used lisp for a longer time (with proper tools like slime)
you would know what i'm talking about. it's completely normal
operation for me to press M-. and jump into the source of the compiler
and look around.


> I also get worried hearing that using a 1.5GB heap causes issues, while
> we routinely use way more than that in Java. I do not want to be
> specific with numbers, cause I don't want trouble at my work.


who's worried? we were only worried that we had *zero* time to debug
stuff. we had 8000 visits and 400000 requests on the *first* day in an
untested ajaxy webapp. i don't say that there were no problems, but
yet after two or three days things were resolved, and on the third day
live data was recorded in the system. so basically we had two days for
"testing". and running siege and httperf is not a replacement for real-
world usage. not to mention that the app wasn't ready on time, so
besides fixing bugs we had to develop some key features while it was
already running.


> There are algorithmic complexities and then there are people+engineering
> complexities.  I get the impression that lots of people here have little
> clue to the people+engineering complexities while they can run circles
> round me about say NLP or continuations for example.


did you read about those 4000 users? all spread around the country?
isn't that cultural complexity to spread passwords for all of them and
collect data with two weeks deadline when some of them had problem
installing firefox?


- attila
From: Pascal Costanza
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <5pdenhFqqgg5U1@mid.individual.net>
··············@gmail.com wrote:
>> That is *not* a lot of libraries. For any real world app one does need
> 
> 
> that was only the list of the important ones, the total is about 60.
> 
> 
>> capabilities. I'd say the 'success' (for some measure) of java owes much
>> to libs, portability, scalability (in terms of processors and memory at
>> least) among other things, even though the core language is poor.
> 
> 
> i agree in that and that's why i'm sorry that the infrastructure
> around library handling is not too great in lisp.

What would be your suggestions for improvement? What are the most 
important entries in your wish list?


Pascal

-- 
My website: http://p-cos.net
Common Lisp Document Repository: http://cdr.eurolisp.org
Closer to MOP & ContextL: http://common-lisp.net/project/closer/
From: Slobodan Blazeski
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194439703.389510.71410@k79g2000hse.googlegroups.com>
On Nov 7, 10:27 am, Pascal Costanza <····@p-cos.net> wrote:
> ··············@gmail.com wrote:
> >> That is *not* a lot of libraries. For any real world app one does need
>
> > that was only the list of the important ones, the total is about 60.
>
> >> capabilities. I'd say the 'success' (for some measure) of java owes much
> >> to libs, portability, scalability (in terms of processors and memory at
> >> least) among other things, even though the core language is poor.
>
> > i agree in that and that's why i'm sorry that the infrastructure
> > around library handling is not too great in lisp.
>
> What would be your suggestions for improvement? What are the most
> important entries in your wish list?

I'll second this, as someone starting with lisp web apps I'm very
interested what lisp lacks in this area. And if it's not a something
secret please give us a site of the java app you're working on.

thanks
Slobodan
>
> Pascal
>
> --
> My website:http://p-cos.net
> Common Lisp Document Repository:http://cdr.eurolisp.org
> Closer to MOP & ContextL:http://common-lisp.net/project/closer/
From: Frank Goenninger DG1SBG
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <lzejf2w835.fsf@pcsde001.de.goenninger.net>
Pascal Costanza <··@p-cos.net> writes:

> ··············@gmail.com wrote:
>>> That is *not* a lot of libraries. For any real world app one does need
>>
>>
>> that was only the list of the important ones, the total is about 60.
>>
>>
>>> capabilities. I'd say the 'success' (for some measure) of java owes much
>>> to libs, portability, scalability (in terms of processors and memory at
>>> least) among other things, even though the core language is poor.
>>
>>
>> i agree in that and that's why i'm sorry that the infrastructure
>> around library handling is not too great in lisp.
>
> What would be your suggestions for improvement? What are the most
> important entries in your wish list?
>
>
> Pascal

Yep - I was also wondering what Attila does need in addition to, say,
CFFI's capabilities in handling libs. CFFI + some system definition
facility like ASDF is what /I/ need. Granted, there are some ASDF
extensions available, but hey, they are *available*.

So - to repeat Pascal:

> What would be your suggestions for improvement? What are the most
> important entries in your wish list?

I am sincerely interested in this. You are using that 60 libs, so you
have first hand experience!

Frank
 
-- 

  Frank Goenninger

  http://www.goenninger.net
  frgo(at)mac(dot)com

  "Don't ask me! I haven't been reading comp.lang.lisp long enough to 
  really know ..."
From: ··············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194561859.854809.55450@t8g2000prg.googlegroups.com>
> Yep - I was also wondering what Attila does need in addition to, say,
> CFFI's capabilities in handling libs. CFFI + some system definition


i didn't mean libs as in shared object's, but libs as in code
libraries.


> facility like ASDF is what /I/ need. Granted, there are some ASDF
> extensions available, but hey, they are *available*.
>
> So - to repeat Pascal:
>
> > What would be your suggestions for improvement? What are the most
> > important entries in your wish list?


asdf extensions do exists, but they are not available "out of the
box" (compare it to the usual argument that everything is doable in
assembly, yet not everything is feasible).

some people will not even reach the point of installing an asdf
extension because (possibly for some other reason) gives up on lisp
before that. i still remember my struggling early days with lisp when
i was creating all those symlinks to .asd files and fighting with
common lisp controller. by now i know that there's a way to use a
lambda that fills the registry scanning some dirs, but still it took
some time to realize it. why? because of the defaults: wrong defaults
guide you to wrong paths, where you stop seeing the big picture
because you trust the defaults.

the two most important extensions i use in asdf are asdf-binary-
locations and asdf-system-connections. they are both extremely small
projects, yet i have to first meet them somehow, then install them,
etc instead of just being part of an integrated library management
tool. and they implement functionality that i think most of us would
like to have.

what i miss most from asdf is version handling. being able to express
constraints beyond the simple fact that x depends on y. then having
support for installing projects from VCS is another thing i miss. and
while we are at it, why is asdf-install a standalone project? why does
such a thing exists as clbuild - a shell script to install lisp
libs...

but these are mostly complaints without much useful thoughts, so i
better hold them back until i invest enough time into more developed
ideas.

but luckily rumour has it that some work is happening on asdf2 (or
anyhow it'll be called).

- attila
From: C Y
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194642830.449660.147750@v3g2000hsg.googlegroups.com>
··············@gmail.com wrote:

> what i miss most from asdf is version handling. being able to express
> constraints beyond the simple fact that x depends on y.

I may be wrong, but I think there is at least basic support for this
in asdf?  Part of the problem is that in order to make sure it would
work asdf would have to enforce the presence of a version number in
any system definition, otherwise at least a few libraries will
probably not include it...

> then having
> support for installing projects from VCS is another thing i miss. and
> while we are at it, why is asdf-install a standalone project? why does
> such a thing exists as clbuild - a shell script to install lisp
> libs...

Good question.  I assume some things in clbuild would be more
difficult to do in Lisp?

To a certain extent, I can sympathize with the difficulties of the
problem - I myself have my own little custom shell scripts to go
through my user level lisp library directories and update.  (I really
need to do something better than that...) For VCS there are at least 4
that are commonly used and the calling conventions for them would have
to be put into the lib (oh, and the calling conventions for external
programs differ between lisp implementations and platforms, IIRC...)

> but these are mostly complaints without much useful thoughts, so i
> better hold them back until i invest enough time into more developed
> ideas.

I agree that it would be nice to have more functionality present in
the default asdf system, but I'm not sure how to accomplish that.

> but luckily rumour has it that some work is happening on asdf2 (or
> anyhow it'll be called).

Interesting - I wasn't aware of that.  Hope it isn't too different
from asdf - the ability to extend asdf has actually come in handy for
me in the past.

Cheers,
CY
From: Frank Goenninger DG1SBG
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <lzbqa1lp34.fsf@pcsde001.de.goenninger.net>
··············@gmail.com writes:

>> Yep - I was also wondering what Attila does need in addition to, say,
>> CFFI's capabilities in handling libs. CFFI + some system definition
>
>
> i didn't mean libs as in shared object's, but libs as in code
> libraries.

Ah yes. That explains a lot. Sorry, I got "libs" as foreign libs
interpreted. My bad.

[snipped some of my own comments]
>
> asdf extensions do exists, but they are not available "out of the
> box" (compare it to the usual argument that everything is doable in
> assembly, yet not everything is feasible).

Yep. Seconded.                                   

> some people will not even reach the point of installing an asdf
> extension because (possibly for some other reason) gives up on lisp
> before that. i still remember my struggling early days with lisp when
> i was creating all those symlinks to .asd files and fighting with
> common lisp controller. by now i know that there's a way to use a
> lambda that fills the registry scanning some dirs, but still it took
> some time to realize it. why? because of the defaults: wrong defaults
> guide you to wrong paths, where you stop seeing the big picture
> because you trust the defaults.

My experience also. I must say if I hadn't had that 10 years Unix
sysadmin experience under my belt I'd also simply scratched my head
over just too many things.

>
> the two most important extensions i use in asdf are asdf-binary-
> locations and asdf-system-connections. they are both extremely small
> projects, yet i have to first meet them somehow, then install them,
> etc instead of just being part of an integrated library management
> tool. and they implement functionality that i think most of us would
> like to have.

Right.                                   

> what i miss most from asdf is version handling. being able to express
> constraints beyond the simple fact that x depends on y. 

Noted as requirement ASDF-R-001. ;-)

> then having
> support for installing projects from VCS is another thing i miss.

ASDF-R-002.

> while we are at it, why is asdf-install a standalone project? why does
> such a thing exists as clbuild - a shell script to install lisp
> libs...

ASDF-R-003.
                                   
> but these are mostly complaints without much useful thoughts, so i
> better hold them back until i invest enough time into more developed
> ideas.

No, no. Just go on - and everybody else also. This is getting really
good. We need it.
                                   
>
> but luckily rumour has it that some work is happening on asdf2 (or
> anyhow it'll be called).
>
> - attila

Thanks!

Now let's go find some time to integrate all those into an ASDF Next
Generation. Brainstoriming mode on over here also...

Frank

-- 

  Frank Goenninger

  frgo(at)mac(dot)com

  "Don't ask me! I haven't been reading comp.lang.lisp long enough to 
  really know ..."
From: C Y
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194796915.701489.75830@v2g2000hsf.googlegroups.com>
On Nov 10, 5:20 pm, Frank Goenninger DG1SBG <·············@nomail.org>
wrote:

> No, no. Just go on - and everybody else also. This is getting really
> good. We need it.

Heh - well, if you insist ;-)

Per various discussions over the years on #lisp, implement a proper
way to have asdf work with versioning.

E.g. be able to specify a particular version of a library, or >= that
version.

Basic idea (credit to various people on #lisp, not me):

1.  define types of version identifications and methods for comparing
each.  E.g.
    (defclass version-id ())  (or maybe type, not class - whatever's
appropriate)
    (defclass version-id-x.y (version-id))
    (defclass version-id-x.y.z (version-id))
    (defclass version-id-x.date (version-id))
    ... (there are a lot of possibilities)

    (defgeneric version-greater-than (version-id))
    (defmethod version-greater-than (version-id-x.y)
         (some comparison code))
    etc...

    This would require some sort of type specification unless a smart
parser can identify versioning types.

2.  epochs (rahul calls them that) - indication and ordering of
earlier version numbering schemes.  Can be safely ignored unless
versioning style changes, in which case an epoch must be defined to
make sure comparisons are unambiguous.  If not present default
assumption would be 1st epoch.

Probably some other details to work out, but that's the general idea.
From: Antony Sequeira
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <ueWdnWz0ms1IsKvanZ2dnUVZ_r2nnZ2d@comcast.com>
··············@gmail.com wrote:
> 
>> And for those who think documentation is not important, think about
>> these guys having to read the source code of all those libs where there
>> is no documentation.
> 
> 
> heh, all of us are one of those guys! :) i _almost never_ read docs!
> exception is when i need to get a bird's view of what the project is
> all about, but those paragraphs i'm looking for then are not those
> that you mean by docs. lisp can be written so that it's better
> readable than a handwritten documentation, and the code itself is
> never outdated!
> 
I have little real Lisp experience. I hope to find you are right.

How about listing more of the libs you use.
I think it was your post that got me to do bit more reading on choices 
for lisp unit test libs.
Thanks for sharing your experience. It helps a lot.
-Antony
From: levy
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194472967.023540.308010@19g2000hsx.googlegroups.com>
On Nov 7, 6:23 am, Antony Sequeira <··················@gmail.com>
wrote:
> And for those who think documentation is not important, think about
> these guys having to read the source code of all those libs where there
> is no documentation.
I'm working with both Java and Lisp at the same time on a weekly basis
for almost two years now, so I think I have some experience to do some
comparison.

One thing I believe why documentation is much less important in lisp
than in Java is that you can write code having better abstractions
(and syntax, once you don't see the parenthesis) which makes
documentation relatively less productive to a level where it might be
more valuable to write code instead. I think it's much harder to read
good Java code than to read good lisp code because there are no easy
ways around the limitations of the Java language and this makes the
syntax really awkward. When I'm saying good here I mean code with good
abstractions without superfluous and error prone redundancy (e.g.
design patterns) and so on...

Just to mention two of my Java favorites:
 - listeners: a terrible solution for (not only gui) change
propagation, who listens to what, when do they start and when do they
finish listening, etc. and still used all over the place
 - code reuse: base64 encoding, there are at least 3 somewhat
different implementations in eclipse non of which is publicly
accessible

In spite of this you are right there are zillions of Java libraries
and lisp could have better library support.

levy
From: gavino
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194477553.020612.83490@e9g2000prf.googlegroups.com>
On Nov 7, 2:02 pm, levy <················@gmail.com> wrote:
> On Nov 7, 6:23 am, Antony Sequeira <··················@gmail.com>
> wrote:> And for those who think documentation is not important, think about
> > these guys having to read the source code of all those libs where there
> > is no documentation.
>
> I'm working with both Java and Lisp at the same time on a weekly basis
> for almost two years now, so I think I have some experience to do some
> comparison.
>
> One thing I believe why documentation is much less important in lisp
> than in Java is that you can write code having better abstractions
> (and syntax, once you don't see the parenthesis) which makes
> documentation relatively less productive to a level where it might be
> more valuable to write code instead. I think it's much harder to read
> good Java code than to read good lisp code because there are no easy
> ways around the limitations of the Java language and this makes the
> syntax really awkward. When I'm saying good here I mean code with good
> abstractions without superfluous and error prone redundancy (e.g.
> design patterns) and so on...
>
> Just to mention two of my Java favorites:
>  - listeners: a terrible solution for (not only gui) change
> propagation, who listens to what, when do they start and when do they
> finish listening, etc. and still used all over the place
>  - code reuse: base64 encoding, there are at least 3 somewhat
> different implementations in eclipse non of which is publicly
> accessible
>
> In spite of this you are right there are zillions of Java libraries
> and lisp could have better library support.
>
> levy

I am curious, from a purely theoretical standpoint....say one did not
use a database.
Say you had a html page generator adn a lisp app to interprest click
on to things liek brose button when someone wanted to upload a file,
that would feed a directy browser to that user allowing her to save a
pdf or otehr document on the unix filesystem in space for thier
account.  The lisp app would take care of saving such files and
referncing them.  Blogs and other collaboration could be handled as a
phantom user (it seems live journal does this).  Chats as well.  With
only lisp and a linux filesystem could you dispense with all of these
modules an simply have the lisp app keep track of everything?
From: Szabolcs  Nagy
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194509232.770448.276180@v3g2000hsg.googlegroups.com>
gavino wrote:
> I am curious, from a purely theoretical standpoint....say one did not
> use a database.
why not? you want to write your own transaction handling layer?

> Say you had a html page generator adn a lisp app to interprest click
> on to things liek brose button when someone wanted to upload a file,
[..]
why wouldn't lisp be able to do that?
or any other sane programming language

and don't tell me you'd write it in php, which is one of the most
horribly designed language out there
From: Frank Buss
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <fqhp601ud5zw.5a6fhth83r0s.dlg@40tude.net>
gavino wrote:

> I am curious, from a purely theoretical standpoint....say one did not
> use a database.

You have to use some kind of database, because you have to save user data
etc. and you don't want to use plain ASCII files, do you?

But there is a Lisp implementation (but not Common Lisp) which has
persistence included, see this tutorial for examples:

http://www.software-lab.de/tut.html

-- 
Frank Buss, ··@frank-buss.de
http://www.frank-buss.de, http://www.it4-systems.de
From: Spiros Bousbouras
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194378395.778699.262990@19g2000hsx.googlegroups.com>
On Nov 1, 5:47 pm, levy <················@gmail.com> wrote:
> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
>
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
>
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.
>
> The entire system is based on opensource software, it runs on multiple
> SBCLs behind an apache load balancer. The data is stored in a
> PostgreSQL server.

Congratulations.

Did bugs in SBCL create any problems like for example
having to rewrite  some piece  of code  which wasn't
working as  it should due  to bugs in  the compiler ?
From: ··············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194380804.216899.108330@19g2000hsx.googlegroups.com>
> Did bugs in SBCL create any problems like for example
> having to rewrite  some piece  of code  which wasn't
> working as  it should due  to bugs in  the compiler ?

i wouldn't put it that way... there are some problems probably caused
by sbcl, but definitely not so big problems that we had to rewrite
parts of the application due to them. most of the suspected problems
are caused by running several threads and sbcl seems to need some more
testing in that regard. we are gathering all the info we can and share
it on sbcl-devel.

- attila
From: gavino
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194400922.677647.144740@y27g2000pre.googlegroups.com>
On Nov 1, 9:47 am, levy <················@gmail.com> wrote:
> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
>
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
>
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.
>
> The entire system is based on opensource software, it runs on multiple
> SBCLs behind an apache load balancer. The data is stored in a
> PostgreSQL server.
>
> We have around 4000 registered users, average 300 online and more than
> 500 at peak times. There are 7000-10000 logins per day.
>
> We use the following Common Lisp libraries. Some of which are ours and
> some of which are built by others. Many thanks for their efforts,
> especially to the SBCL developers!
>
>   - Data storage/persistence
>     - cl-perec
>     - cl-rdbms
>     - cl-postgres (postmodern)
>
>   - Web GUI
>     - cl-dwim (the umbrella project)
>     - UCW
>     - yaclml
>     - parenscript
>     - rfc*
>
>   - PDF exporting
>     - cl-pdf
>     - cl-typesetting
>
>   - Unit testing
>     - stefil
>     - SLIME and emacs with customizations
>
>   - MOP
>     - computed-class
>     - contextl
>
>   - Misc
>     - bind
>     - alexandria
>     - iterate
>     - cl-def
>     - cl-l10n
>     - cl-graphviz
>     - cl-serializer
>     - defclass-star
>     - local-time
>     - cffi
>     - babel
>     - verrazano
>
> and more...
>
> Main software components:
>   - Ubuntu linux 6.10
>   - PostgreSQL 8.2
>   - SBCL 1.0.10
>   - Apache2 (due to various headaches soon to be replaced by a lisp
> based load balancer)
>
> Hardware:
>   - 10 nodes with two dual core x86-64 CPUs summing up to 40
> cores altogether, each node has 3 Gbytes ram.
>   - Currently we use one of the nodes as database server but soon it
> will
> be replaced with another server with 2 or 4 quad core x86-64 CPUs, 300
> GByte disk in raid 5 and 16 GBytes ram.
>
> Some people say Lisp is dead but seems like there's life after
> death... ;-)

how does lisp help to get things done vs say php?
From: ············@gmail.com
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194629958.047558.234050@19g2000hsx.googlegroups.com>
On Nov 1, 11:47 am, levy <················@gmail.com> wrote:
> We are a small group of friends (3-5) who left the "professional"
> programming scene about one and a half year ago to start our own Lisp
> business.
>
> In July, 2007 we started working on a project for the Hungarian
> government with a very tight deadline. In mid October the system
> showed first life signs and now we can say it is extensively used by
> thousands of users. In spite of the short timeframe and the remaining
> glitches it was a surprise even for us that the system is able to
> handle such a load.
>
> Unfortunately the application is not publicly accessible, so we can't
> provide a URL, but we can give some of the details. It is used for
> gathering data from the Hungarian communes to help the budget
> planning.
>
> The entire system is based on opensource software, it runs on multiple
> SBCLs behind an apache load balancer. The data is stored in a
> PostgreSQL server.
>
> We have around 4000 registered users, average 300 online and more than
> 500 at peak times. There are 7000-10000 logins per day.
>
> We use the following Common Lisp libraries. Some of which are ours and
> some of which are built by others. Many thanks for their efforts,
> especially to the SBCL developers!
>
>   - Data storage/persistence
>     - cl-perec
>     - cl-rdbms
>     - cl-postgres (postmodern)
>
>   - Web GUI
>     - cl-dwim (the umbrella project)
>     - UCW
>     - yaclml
>     - parenscript
>     - rfc*
>
>   - PDF exporting
>     - cl-pdf
>     - cl-typesetting
>
>   - Unit testing
>     - stefil
>     - SLIME and emacs with customizations
>
>   - MOP
>     - computed-class
>     - contextl
>
>   - Misc
>     - bind
>     - alexandria
>     - iterate
>     - cl-def
>     - cl-l10n
>     - cl-graphviz
>     - cl-serializer
>     - defclass-star
>     - local-time
>     - cffi
>     - babel
>     - verrazano
>

Where is the library "babel"?  What does it do?  Can I have access to
it?  cl-serializer depends on it.

Eric
From: Maciej Katafiasz
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <fh2ha9$k4n$2@news.net.uni-c.dk>
Den Fri, 09 Nov 2007 17:39:18 +0000 skrev ericwnormand:

> Where is the library "babel"?  What does it do?  Can I have access to
> it?  cl-serializer depends on it.

http://common-lisp.net/~loliveira/soc07/babel/

Cheers,
Maciej

PS. Consider cutting down the text you quote a bit if all you want is to 
ask a simple question.
PS2. When should we expect the next lispcast? :)
From: Eric
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194655697.107635.290600@19g2000hsx.googlegroups.com>
On Nov 9, 2:52 pm, Maciej Katafiasz <········@gmail.com> wrote:
> Den Fri, 09 Nov 2007 17:39:18 +0000 skrev ericwnormand:
>
> http://common-lisp.net/~loliveira/soc07/babel/

Thanks

> PS. Consider cutting down the text you quote a bit if all you want is to
> ask a simple question.

Oops, sorry.

> PS2. When should we expect the next lispcast? :)

Barring any disasters, this weekend.

Eric
From: Sacha
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <X_nZi.200446$tV.11200765@phobos.telenet-ops.be>
Eric wrote:
> On Nov 9, 2:52 pm, Maciej Katafiasz <········@gmail.com> wrote:
>> Den Fri, 09 Nov 2007 17:39:18 +0000 skrev ericwnormand:
>> PS2. When should we expect the next lispcast? :)
> 
> Barring any disasters, this weekend.
> 
> Eric

Yay !

Sacha
From: Eric
Subject: Re: We are happy to announce: our first Lisp production system went online
Date: 
Message-ID: <1194804403.478515.29870@50g2000hsm.googlegroups.com>
On Nov 10, 1:53 pm, Sacha <····@address.spam> wrote:
> >> When should we expect the next lispcast? :)

> > Barring any disasters, this weekend.

> Yay !


LispCast Episode 4 is up now.

http://www.lispcast.com

Eric