Ok, I've written my first, from the ground up, Comanche server module.

Yay.

It was to test out just using XML and HTTP (as opposed to some RPC over
them) for client/server apps. But the actually application is pretty
neat. I'm going to comment it up and post it maybe tomorrow (hah!).

(It's a chat server which uses XML as the format, but can GET and POST
messages. What's *really* cool is that I hooked it up to IRC, so each
"send" on an irc channel became accessible from HTTP. It already has very
simple clients in Perl and Java (er...better get that squeak one
done!). And they were quite trivial to write, btw (most of the trouble was
remember what protocals we agreed on ;)). Plans include: simple
persistence, HTML display and posting, two-way gateway to IRC (that's the
cool one :)), etc.)

I do have a comment, though maybe this will come clear as documentation
emerges: There seems an awful lotta a ways to do things. Too many, at
least without clear guidelines. I was jumping all over. At some point, I
think a nice simplification (a la the great future morphic clean up) would
be in order.

Two questions:

1) I don't quite understand the url walking stuff. And, in particualrly, I
was wondering why the Url classes weren't used? It seems to me that
they're the right thing.

2) Is there a straightforward way to get the "body" of a POST that *isn't*
form field encoded? Say, if it has a content-type of text/xml,
perchance? I tried upToEnding the stream, but that didn't seem to
work. #body (to contrast with #headers) seems like a good thing to have,
but I'd take a #contents.

Of course, the particular server I was using was Pre-FourStrongWinds. So
maybe these are resolved there. Dang!

Cheers,
Bijan.

Reply via email to