[Lispweb] Araneida

[Cyrus Harmon]

My experience is that the net.aserve.client part of the latest CVS 
version of paserve and 0.8.21.50 mostly work but that when they fail 
the symptom is the quiet, sporadic insertion of characters (well, a 
single character, but it happens somewhere between 0 and 10 or so 
times) into streams from certain websites. I probably wouldn't have 
noticed this except for the fact that I got "lucky" and one of the 
characters plopped itself into an XML element tag. I'm getting off the 
topic of lisp-based web servers and should take this to the paserve 
list, but I wanted to address Rudi's claim of workingness. Attempting 
to debug this code proved to be a huge pain. It took me about 30 
minutes to get trivial-http running instead of the paserve stuff for my 
needs. I'm going to hand this one off to the paserve folks and hope I 
don't have to use it soon (BTW, the problem also occurs on OpenMCL 
0.14.3, just less frequently). Brian may not have a special spot in his 
heart for web code written in C; personally, I hate debugging stuff 
with if* all over the place.

Cyrus

On Apr 21, 2005, at 8:12 AM, Rudi Schlatte wrote:

>
> On 21. Apr 2005, at 14:51, Brian Mastenbrook wrote:
>> paserve, on the other hand, depends on a huge library designed to 
>> turn the host lisp into an approximation of allegro. I don't have any 
>> problems with the paserve API as such, but it's worth comparing the 
>> thousands of lines of code in acl-compat to the compat/ directory in 
>> araneida, and determining whether the resulting complexity (and 
>> fragility, when the host lisp changes: I don't think paserve works in 
>> SBCL 0.8.21, according to reports on IRC) is worth it.
>
> I haven't checked before sending this mail, but I think paserve CVS 
> (and the asdf-installable version) should work on new sbcls; the 
> released version lags behind, and I won't be able to release before 
> Amsterdam.
>
> Re acl-compat: I concur, it's basically a hack.  It would be nice (tm) 
> to have a lighter compatibility layer, but I try to alter the upstream 
> code as little as possible, because I would start by exorcising every 
> trace of if* from it and would lose the ability to merge upstream 
> changes easily.  Coding style aside, I like AllegroServe but know that 
> I would not keep up with maintaining it in the long run, so acl-compat 
> is what keeps paserve alive.
>
> (For those who haven't looked at AllegroServe: it's a full-featured 
> HTTP 1.1-compliant webserver that can serve static and dynamic 
> content, serve as a proxy, execute CGI-Scripts (in theory, it's not 
> supported by acl-compat at the moment) and has a client module.)
>
> Rudi
> _______________________________________________
> Lispweb mailing list
> Lispweb at red-bean.com
> http://www.red-bean.com/mailman/listinfo/lispweb