Hello,
I've made a website in PHP, and in five years, the code has gotten
fat, complicated and ugly. And since I'll need to revamp it anyway (to
allow non-programmers to change all the pages, wiki-like), I intend to
restart from scratch, some time around spring next year.
I intended to do it in PHP again.
I just discovered LISP, and it occured to me that it'd be a good idea
to choose that language instead -- if only to get the refreshing taste
of a completely different language.
Is that a good idea?
I mean, PHP is good at creating HTML (since it was created for that
goal), has good portability and easy deployment (a simple rsync is
sufficient to make the code I developed on my Windows PC, work
perfectly on the Linux web server), and good integration in Apache.
Will I find the same qualities in Lisp?
Thanks in advance for your advice.
Fabien LE LEZ wrote:
> Hello,
>
> I've made a website in PHP, and in five years, the code has gotten
> fat, complicated and ugly. And since I'll need to revamp it anyway (to
> allow non-programmers to change all the pages, wiki-like), I intend to
> restart from scratch, some time around spring next year.
> I intended to do it in PHP again.
>
> I just discovered LISP, and it occured to me that it'd be a good idea
> to choose that language instead -- if only to get the refreshing taste
> of a completely different language.
>
> Is that a good idea?
>
Lisp can empower you to easily create and extend complex functionality
as no other language can ... ! (Well, atleast I think so)
> I mean, PHP is good at creating HTML (since it was created for that
> goal),
HTTP/HTML is a very simple protocol, and there are several Lisp
libraries/modules which already exist to kickstart leverage for your
specific needs. Although I am not familiar with them, you can take your
pick from one of the several listed here
http://www.cliki.net/web
> has good portability and easy deployment (a simple rsync is
> sufficient to make the code I developed on my Windows PC, work
> perfectly on the Linux web server), and good integration in Apache.
> Will I find the same qualities in Lisp?
I would think that *most* Lisp code should be portable across OS, and
web servers. But again even all programs written in Java are not 100%
portable across OSes etc.
>
> Thanks in advance for your advice.
HTH
Alok
On 20 Aug 2006 15:40:22 -0700, "Alok" <···········@gmail.com>:
>> I mean, PHP is good at creating HTML (since it was created for that
>> goal),
>
>HTTP/HTML is a very simple protocol
I don't care much about HTTP (I let Apache handle 99% of the work),
but HTML is indeed simple -- it's just text.
The thing is, PHP is good at manipulating text (and at doing the kind
of manipulation we do when creating HTML code). But it doesn't scale
very well -- after a few thousand lines of code, it kinda begins to be
a mess.
OTOH, the other language I know well, C++, will be at ease with a
10-million-line project, but just sucks at string manipulations. And
ain't easy to deploy, since it needs compiling every time there's a
modification in the code.
In 1995, Paul Graham <http://www.paulgraham.com/avg.html> made his
website in Lisp. Would he have done so in 2006, now that PHP is there?
Fabien LE LEZ wrote:
> On 20 Aug 2006 15:40:22 -0700, "Alok" <···········@gmail.com>:
>
> >> I mean, PHP is good at creating HTML (since it was created for that
> >> goal),
> >
> >HTTP/HTML is a very simple protocol
>
> I don't care much about HTTP (I let Apache handle 99% of the work),
> but HTML is indeed simple -- it's just text.
For building dynamic HTML content, access to HTTP layer is required,
especially request/response attributes, message headers etc.. Any good
web framework should have objects and routines for access to it.
Although this may not be good bulk of a web famework, it is an
important chunk.
Alok.
HTML represents a hierarchical data structure, easily manipulated as a
tree using linked lists, a natural data structure for processing in
Lisp. The text file format is used for storage and transmission. It is
however not generally efficient to process html/xml in its' text
representation - if anything this is where languages like PHP and Perl
go wrong and make life hard and where Lisp gains.
>>>>> "Fabien" == Fabien LE LEZ <········@gramster.com> writes:
Fabien> On 20 Aug 2006 15:40:22 -0700, "Alok"
Fabien> <···········@gmail.com>:
>>> I mean, PHP is good at creating HTML (since it was created for
>>> that goal),
>> HTTP/HTML is a very simple protocol
Fabien> I don't care much about HTTP (I let Apache handle 99% of
Fabien> the work), but HTML is indeed simple -- it's just text.
Fabien> The thing is, PHP is good at manipulating text (and at
Fabien> doing the kind of manipulation we do when creating HTML
Fabien> code). But it doesn't scale very well -- after a few
Fabien> thousand lines of code, it kinda begins to be a mess.
Fabien> OTOH, the other language I know well, C++, will be at ease
Fabien> with a 10-million-line project, but just sucks at string
Fabien> manipulations. And ain't easy to deploy, since it needs
Fabien> compiling every time there's a modification in the code.
Fabien> In 1995, Paul Graham <http://www.paulgraham.com/avg.html>
Fabien> made his website in Lisp. Would he have done so in 2006,
Fabien> now that PHP is there?
--
John Williams
Fabien LE LEZ <········@gramster.com> writes:
>
> The thing is, PHP is good at manipulating text (and at doing the kind
> of manipulation we do when creating HTML code). But it doesn't scale
> very well -- after a few thousand lines of code, it kinda begins to be
> a mess.
There are Lisp templating engines (e.g. Webactions and CL-EMB), and also
libraries to generate HTML from Lisp (e.g. HTMLGEN). An example might
look like this:
(html
(:html
((:head lang "en_GB")
(:title "A Little Page"))
(body
(:h1 "A Little Page")
(:p "Some text goes here))))
The templating engines would be more like what you're used to with PHP,
but the code-generated option can be very powerful.
> OTOH, the other language I know well, C++, will be at ease with a
> 10-million-line project, but just sucks at string manipulations. And
> ain't easy to deploy, since it needs compiling every time there's a
> modification in the code.
Common Lisp can help you with both of those.
--
Robert Uhl <http://public.xdi.org/=ruhl>
NT is a one-legged cow, but even a one legged cow is fast when it's
got 160+ rockets strapped to it. --Nick Manka
But that's not that impressive if all you can make it do is go around
in circles. --Darrell Fuhriman
Robert Uhl <·········@nospamgmail.com> wrote:
> The templating engines would be more like what you're used to with PHP,
> but the code-generated option can be very powerful.
And you can mix both.
--
Web (en): http://www.no-spoon.de/ -*- Web (de): http://www.frell.de/
> The thing is, PHP is good at manipulating text (and at doing the kind
> of manipulation we do when creating HTML code). But it doesn't scale
> very well -- after a few thousand lines of code, it kinda begins to be
> a mess.
I have to disagree here. Scalability is generally a design issue and
not a language issue. Java, C#, Python, Ruby and even Lisp programs
can get pretty nasty if there is a bad design.
Fabien LE LEZ <········@gramster.com> writes:
> Is that a good idea?
Yes, it's a good idea. But beware: Most frameworks assume that you are
able to install/run software on your webserver, so if you have only
access to a cheap webhosting with only static HTML and a bit PHP
and/or CGI allowed, you have to look a bit closer (Lisp webapps may
run via CGI, but as a seperate process that communicates via socket
with Apache the performance is better).
Some links to get startet:
http://www.cl-user.net/asp/tags/web
http://weitz.de/tbnl/
http://www.weitz.de/cl-who/
http://common-lisp.net/project/ucw/
http://www.cliki.net/ucw Tutorial
http://common-lisp.net/project/cl-emb/
http://www.cliki.net/parenscript
http://clsql.b9.com/
--
Stefan.
On Mon, 21 Aug 2006 12:00:30 +0200, Stefan Nobis <······@gmx.de>:
>But beware: Most frameworks assume that you are
>able to install/run software on your webserver,
I've got root access to the server, so it should be fine :-)
>but as a seperate process that communicates via socket
>with Apache the performance is better).
Well, well...
I'll have to learn a new programming paradigm (LISP itself), a new
HTML-handling paradigm (tree instead of text), and a new
web-programming paradigm (one big application that handles every
requests, instead of one little script per request).
That should keep me busy for a little while...
Fabien LE LEZ <········@gramster.com> writes:
> I've got root access to the server, so it should be fine :-)
If you got control of the server host you can even run a web-server
implemented in Common Lisp, like AllegroServe/WebActions.
AllegroServe/WebActions, CLSQL and SLIME is a very productive
environment.
Petter
--
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?
Fabien LE LEZ <·······@edulang.com> wrote:
+---------------
| Stefan Nobis <······@gmx.de>:
| >but as a seperate process that communicates via socket
| >with Apache the performance is better).
|
| Well, well...
| I'll have to learn a new programming paradigm (LISP itself), a new
| HTML-handling paradigm (tree instead of text), and a new
| web-programming paradigm (one big application that handles every
| requests, instead of one little script per request).
+---------------
You don't need "one big application" per se: One of the useful
features of Common Lisp is that you can compile & load code into
a running program, so you can continue to write separate "web pages"
of code, and have them auto-loaded [and auto-recompiled if they
change!] into the Lisp application server "on demand" [with "JIT"
compilation]. My own LHP (Lisp-Handled Pages) infrastructure does it
that way. Here are a couple of simple examples:
http://rpw3.org/hacks/lisp/minimal.lhp
http://rpw3.org/hacks/lisp/appsrv-demo.lhp
Many other people have done similar things, I sure...
-Rob
-----
Rob Warnock <····@rpw3.org>
627 26th Avenue <URL:http://rpw3.org/>
San Mateo, CA 94403 (650)572-2607
On Tue, 22 Aug 2006 07:38:49 -0500, ····@rpw3.org (Rob Warnock):
> http://rpw3.org/hacks/lisp/minimal.lhp
> http://rpw3.org/hacks/lisp/appsrv-demo.lhp
Well, now I know that LISP is a funny language:
look at all those :p smileys...
Fabien LE LEZ wrote:
> On Tue, 22 Aug 2006 07:38:49 -0500, ····@rpw3.org (Rob Warnock):
>
> > http://rpw3.org/hacks/lisp/minimal.lhp
> > http://rpw3.org/hacks/lisp/appsrv-demo.lhp
>
> Well, now I know that LISP is a funny language:
> look at all those :p smileys...
grins-p
Le Mon, 21 Aug 2006 00:17:35 +0200, Fabien LE LEZ a écrit :
> Is that a good idea?
As I'm currently doing it, I think it is an excellent one!
> Will I find the same qualities in Lisp?
Not exactly. As it was said, you won't find the same ease of
installation at first, because PHP is now the de facto standard and is
nearly universally available.
But you will find many more qualities: the 'canonical' way to handle Web
content with Lisp seems to have an Lisp image constantly running and
serving that content. That removes a great deal of overhead, both in run
time and code complexity.
The use of lambda functions created by macros and other functions and
the use of functions as first-class objects also gives a great deal of
flexibilty tou your app.
As far as specific tools are concerned, I was very pleased to use
Araneida, because it make it simple to use an Apache front-end as an
HTTP cache, thus increasing speed while decreasing the load on the
server, and Elephant, because it was needed to do hardly more than
adding an options to one of my classes to make it persistent accross
times when the image is running (of course, persistence accross HTTP
request is also provided by the fact that the image is running in
between).
And last bu not least, the mapping between HTML and S-expression make it
a real pleasure to build the structure of the page, and it becomes very
easy to make that structure more dynamic, without really losing in
readability.
Happily,
Nowhere man
--
···········@levallois.eu.org
OpenPGP 0xD9D50D8A
On Mon, 21 Aug 2006 15:34:48 +0200, Pierre THIERRY
<···········@levallois.eu.org>:
>But you will find many more qualities: the 'canonical' way to handle Web
>content with Lisp seems to have an Lisp image constantly running and
>serving that content.
Does this work well when several sites, made by several (independant)
programmers, are hosted on the same machine (with Apache's virtual
domains)?
Le Thu, 24 Aug 2006 20:26:24 +0200, Fabien LE LEZ a écrit :
>>But you will find many more qualities: the 'canonical' way to handle
>>Web content with Lisp seems to have an Lisp image constantly running
>>and serving that content.
> Does this work well when several sites, made by several (independant)
> programmers, are hosted on the same machine (with Apache's virtual
> domains)?
I asked myself the question. I was told, on IRC I think, that this was
the way Viaweb was built. And it dealt with a fair amount of sites,
though it is not said on how many servers...
If it has to be done with many sites, I think a solution where most of
the core code of Lisp is shared, as in ECL, could be used. Then, each
image only weights it's own specific code, which can be pretty small.
With compiled code and proper caching, this could prove to be very easy
to deal with for the server, I think.
It would be interesting to see how this can scale up, and if it would be
possible to rather have a single Lisp image do everything and keep code
of different sites in isolated contexts (à la eval with environment like
in Scheme, maybe).
Doubtfully,
Nowhere man
--
···········@levallois.eu.org
OpenPGP 0xD9D50D8A
On Thu, 24 Aug 2006 22:22:37 +0200, Pierre THIERRY
<···········@levallois.eu.org>:
>I asked myself the question. I was told, on IRC I think, that this was
>the way Viaweb was built. And it dealt with a fair amount of sites,
>though it is not said on how many servers...
AFAIK, Viaweb was one application, made by a handful collaborating
programmers.
Le Thu, 24 Aug 2006 22:44:07 +0200, Fabien LE LEZ a écrit :
> AFAIK, Viaweb was one application, made by a handful collaborating
> programmers.
It provided the ability to create an online store automatically, IIUC. I
suppose each store had it's own image.
Doubtfully,
Nowhere man
--
···········@levallois.eu.org
OpenPGP 0xD9D50D8A
Fabien LE LEZ <········@gramster.com> writes:
> Does this work well when several sites, made by several (independant)
> programmers, are hosted on the same machine (with Apache's virtual
> domains)?
Allegroserve supports virtual hosts. You can also use Apache as a
frontend by redirecting to a specific port number on the same host (or
a different host for that matter).
Petter
--
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?
On 25 Aug 2006 01:18:15 +0200, Petter Gustad
<·············@gustad.com>:
>> Does this work well when several sites, made by several (independant)
>> programmers, are hosted on the same machine (with Apache's virtual
>> domains)?
>
>Allegroserve supports virtual hosts.
But that doesn't mean it can completely separate the Lisp code for
www.foo.com and the Lisp code for www.bar.com, so that it's impossible
for the former to do anything on the latter.
Ideally, there should be no way for [the Lisp code of] one website to
know that other websites are hosted on the same machine.
Note that it's not really a question of virtual domains: typically,
you don't want http://example.com/~foo and http://example.com/~bar to
interfere in any way.
Fabien LE LEZ <········@gramster.com> writes:
> On 25 Aug 2006 01:18:15 +0200, Petter Gustad
> <·············@gustad.com>:
>
> >> Does this work well when several sites, made by several (independant)
> >> programmers, are hosted on the same machine (with Apache's virtual
> >> domains)?
> >
> >Allegroserve supports virtual hosts.
>
> But that doesn't mean it can completely separate the Lisp code for
> www.foo.com and the Lisp code for www.bar.com, so that it's impossible
No, the opposite. You can run multiple instances of Allegroserve
talking to different ports (or even on separate hosts) which is mapped
by Apache if you want to do that.
> for the former to do anything on the latter.
> Ideally, there should be no way for [the Lisp code of] one website to
> know that other websites are hosted on the same machine.
However, I think it's a benefit to share much of the code for several
sites. Quite often they share much functionality and with the
condition system of Common Lisp you can make a very robust system.
> Note that it's not really a question of virtual domains: typically,
> you don't want http://example.com/~foo and http://example.com/~bar to
> interfere in any way.
If you worry about the interference between two web servers you might
as well use multiple hosts.
Petter
--
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?
Le Fri, 25 Aug 2006 12:56:19 +0200, Petter Gustad a écrit :
>> But that doesn't mean it can completely separate the Lisp code for
>> www.foo.com and the Lisp code for www.bar.com, so that it's
>> impossible
> No, the opposite. You can run multiple instances of Allegroserve
> talking to different ports (or even on separate hosts) which is mapped
> by Apache if you want to do that.
The problem is: if the code is run in the same Lisp image, it could
avoid the other site working as expected: modify packages, functions,
etc. I'm not sure this is acceptable.
Quicky,
Nowhere man
--
···········@levallois.eu.org
OpenPGP 0xD9D50D8A
On Fri, 25 Aug 2006 22:06:30 +0200, Pierre THIERRY
<···········@levallois.eu.org>:
>The problem is: if the code is run in the same Lisp image, it could
>avoid the other site working as expected: modify packages, functions,
>etc. I'm not sure this is acceptable.
Definitely not.
In a more static language, like C++, there's no problem: some code can
be shared (using static- ou dynamic-linked libraries, or directly by
sharing source code), but there are two copies of the code in memory,
and one program can't change the code (or even the data) of another.
Pierre THIERRY <···········@levallois.eu.org> writes:
> Le Fri, 25 Aug 2006 12:56:19 +0200, Petter Gustad a �crit�:
> >> But that doesn't mean it can completely separate the Lisp code for
> >> www.foo.com and the Lisp code for www.bar.com, so that it's
> >> impossible
> > No, the opposite. You can run multiple instances of Allegroserve
> > talking to different ports (or even on separate hosts) which is mapped
> > by Apache if you want to do that.
>
> The problem is: if the code is run in the same Lisp image, it could
> avoid the other site working as expected: modify packages, functions,
> etc. I'm not sure this is acceptable.
The different instances of Allegroserve can run in separate lisp
processes. If you consider the developers as potential malicious
hackers you might want to have them running on separate hosts.
Petter
--
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?
On 27 Aug 2006 20:06:57 +0200, Petter Gustad
<·············@gustad.com>:
> If you consider the developers as potential malicious
>hackers
Nope, programmers that can make mistakes.
In a sense it's worse: you might have a pretty good idea of what a
cracker will attempt, but you can't predict what a mistake will do.
Fabien LE LEZ <········@gramster.com> writes:
> On 27 Aug 2006 20:06:57 +0200, Petter Gustad
> <·············@gustad.com>:
>
> > If you consider the developers as potential malicious
> >hackers
>
> Nope, programmers that can make mistakes.
You should probably keep them on different servers since you never
know if they type rm -fr / by mistake...
Petter
--
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?
Le Sun, 27 Aug 2006 22:09:39 +0200, Petter Gustad a écrit :
> You should probably keep them on different servers since you never
> know if they type rm -fr / by mistake...
Why the hell would you give them authority to do that? They would be
confined (as far as UNIX can confine) in their shell account with no
right to do anything unneeded outside it.
At best, you would make them run their image in a chroot or a BSD jail.
Securely,
Nowhere man
--
···········@levallois.eu.org
OpenPGP 0xD9D50D8A
Pierre THIERRY <···········@levallois.eu.org> writes:
> Le Sun, 27 Aug 2006 22:09:39 +0200, Petter Gustad a �crit�:
>> You should probably keep them on different servers since you never
>> know if they type rm -fr / by mistake...
>
> Why the hell would you give them authority to do that? They would be
> confined (as far as UNIX can confine) in their shell account with no
> right to do anything unneeded outside it.
>
> At best, you would make them run their image in a chroot or a BSD jail.
Once upon a time, a debugged and tested a script under a user
account. It worked very well. But for production, it had to run under
root. Unfortunately, it contained a: rm -rf $DIR/$SUBDIR, and doubly
unfortunately, when I ran it the first time as root, both DIR and
SUBDIR were empty...
In the end, I patched rm....
http://www.informatimago.com/develop/patches/index.html
--
__Pascal Bourguignon__ http://www.informatimago.com/
NOTE: The most fundamental particles in this product are held
together by a "gluing" force about which little is currently known
and whose adhesive power can therefore not be permanently
guaranteed.
On Sun, 27 Aug 2006 22:03:46 +0200, Pascal Bourguignon
<···@informatimago.com>:
>But for production, it had to run under root.
That's a very common problem on Linux: far too many scripts and
programs run as root.
The other (and linked) problem is that there doesn't seem to be a
working log-structured filesystem...
OT, so fu2 poster.
(Yes, Usenet can be as cryptic as Lisp ;-p )
Pierre THIERRY <···········@levallois.eu.org> writes:
> At best, you would make them run their image in a chroot or a BSD
> jail.
I think I've misunderstood. You are not planning to have your trusted
employees working on a multiple site web application, but rather
setting up some kind of web-hosting service...
Petter
--
A: Because it messes up the order in which people normally read text.
Q: Why is top-posting such a bad thing?
A: Top-posting.
Q: What is the most annoying thing on usenet and in e-mail?
Fabien LE LEZ <········@gramster.com> writes:
> I just discovered LISP, and it occured to me that it'd be a good idea
> to choose that language instead -- if only to get the refreshing taste
> of a completely different language.
>
> Is that a good idea?
Can't see why not. I work on a commercial app written in lisp
whose UI is written in HTML. It's very easy to extend and do
nice things in; and your response functions can speak to
objects tied to a DB seamlessly.
We're now rewriting it to make heavy use of javascript, and
it's turning out even cooler and cleaner this time around.
[Of course, that might be because I _finally_ know what I'm
doing...]
--ap
As it happens, I just did one of these for fun just today!
My elderly dad can't cope with even slightly complex mailers, and only
needs to talk to my sister, my brother, and me. So I built us a family
mail system:
http://nostoc.stanford.edu/jeff/temp/dadsrv.lisp
It's probably not the cleanest code in the world, but it does the job
in ~150 lines of code. This took an afternoon to write, debug, and
setup as a server that my family can now use (and already has!)
(The only functionality that isn't standard are pretty-time and
pretty-date which come from my utils. Note that I've replaced some of
the identifying strings with XXX for my own safety.)
Cheers,
'Jeff