From: Rainer Joswig
Subject: Re: Dead or not,
Date: 
Message-ID: <joswig-ya023180003107970015560001@news.lavielle.com>
In article <·····································@gairsay.aiai.ed.ac.uk>,
····@aiai.ed.ac.uk wrote:

> But a kind of bottom line is this:
> 
>   I think Harlequin's Lisp implementors ought to be able to make
>   Common Lisp be able to deliver "applications" that are just as
>   efficient as those delivered by Dylan.

I personally think that Common Lisp's problems
(footprint of applications, application creation process, ...)
are fixable, should be addressed by the vendors and
customers should ask for that. The more people are starting
to develop real applications (atleast on the Mac with
MCL I can see these people) the more vendors will need
to support them.

Speed efficiency is not a real problem nowadays (IMHO). Bigger problems
are:

a)  support for platform specific technologies (this is
    currently really bad on the PC, I wonder how Dylan will do here???)
b)  space efficiency of the basic Lisp runtime (can be addressed
    by shared libraries so that multiple applications can reuse
    the libraries they need - this will generate a problem with different
    library versions. Shared Libraries should also come in a special
    application deployment version).
c)  space efficiency of the running application's RAM requirements
d)  startup time (important if you like to write traditional CGIs). 
e)  universality
    - where can I deploy my code? Can I deploy it on a Sony Playstation?
      Why not? Can I write a Netscape plugin?
    - can it be incorporated in foreign code?
f)  critical mass of developers that "standards" will/can be supported:
    TCP/IP, HTTP, CORBA, Active X, Open GL, CGI, Windows GUI, ...

Currently there is not that much need to develop Lisp applications
specially for Unix. Windows NT PCs have more horsepower
than needed (multiple processors, >200Mhz Pentiums, hundreds
of MBs of RAM, GB disks, fast ethernet, accelerated graphics,
relatively stable OS, ...). Prices are going done fast.
We just need some people with the will to use what's here.

Will Dylan be just a cool development system? This may not be enough.
For my purposes CL is already here. I would have to wait for things
like Dylan-HTTP or write it myself.

-- 
http://www.lavielle.com/~joswig/