[Lispweb] Mod_lisp 2.0 released

Craig Brozefsky craig at red-bean.com
Mon Jun 18 00:59:49 CDT 2001


"Andrew K. Wolven" <awolven at redfernlane.org> writes:

> So the point that you are trying to make is that your employer had
> you dinking around with inferior java technology in order to support
> a mostly inferior web server in order to support an inferior
> operating system in order to support yet another mostly inferior
> programmer who calls button-pushing a 'science'.  (All of this
> having nothing to do with lisp or web except to justify a system
> which could have just used fastcgi or jserve in the first place if
> you must sell buzztalk.)

With regards to the use of mod_jserv, which I can only assume is what
you are referring to above, it saved us prolly a month in development
time to use that as opposed to developing a webserver in CL that would
do the job sufficiently.  At that time Allegroserve was not available
(and I would like in the future to have Allegroserve acting as a
frontend to imho).

The decision has nothing to do with "buzztalk" or being Java ready.
It had to do with how we could get our CL based application into
production in a manner capable of supporting our client's expected
loads as quickly as possible.  

There is nothing in our application that precludes the move to an
integral webserver.  To this day however there is no business case,
and really no technical case, that would justify the time it would
take to move there.  I'm just not understanding what you are getting
at by making such a large deal about such a minor part of any web
application.

> Consider this common abbreviation of one of my employers: AIMD,
> Aviation Intermediate Maintenance Depot.  The only Tomcat technology
> AIMD bothers to support is F-14 Tomcat.  In this environment, you
> may find yourself facing both a court martial and the wrath of a
> widow if money makes you complacent and you cannot do your homework.

What's your point?  What should I do, conceed that your bluster is
sufficient because of who pays you?

> Of course, that example is way extreme, let us consider a less
> mission critical application environment: an Online store.  Do you
> want to rewrite the entire system from scratch to make the jump from
> 100 customers a month to 100,000?  It's up to you, it is your time,
> your money and your carpal tunnel syndrome.

I fail to see how moving from a shim to an integral webserver
constitutes rewriting the entire application.

Also, considering the volume of online stores using the "shim" model,
such as the Apple store, I'm not sure you have made a strong
connection between the use of the shim and scalability in these real
world applications.

To turn the tides here.  I think you'll find that unless you're using
a process-level load-balancer (something that these "shims" often do)
that the MP capabilities of the available CL implementations will
become your bottleneck.  In fact, for our application, the MP support
in x86 CMUCL is more of an issue than the overhead per HTTP connection
of going thru a shim.

Also, in terms of the time spent going thru a complete
request-response loop, the overhead of using a shim is largely
inconsequential for our application.  The time spent pulling things
from database, or otherwise processing requests dwarfs it.

-- 
Craig Brozefsky                             <craig at red-bean.com>
                                  http://www.red-bean.com/~craig
"Indifference is the dead weight of history." -- Antonio Gramsci



More information about the lispweb mailing list