[Hessianphp-development] RE: couple questions

[<veg...@gm...>]

Hi Jonathan

Sorry for the delay, and thanks. Now about your questions:

-- Curl: 
Curl by itself does seem to be faster than using php own socket
functions, just because the extra effort of parsing HTTP headers that
you *have* to do in raw php, altough the performance gain is almost
insignificant.
Recently I was performing a few tests over CURL to see if it's
possible to use it in the data sending mechanism of HessianPHP.
After some time messing with curl_setopt I was able to send binary
data to both HessianPHP and Hessian Java, but it has some really ugly
caveat. The only way I can make a Hessian service understand binary
data coming from PHP is to simulate a file upload(!). Curl will accept
a file descriptor (resource) for a file upload, but only it it point
to a real file. I tried to simulate it using php streams (
stream_wrapper_register() ), that can act as file descriptors without
being a real file, but Curl keeps reject it. So the only way was to
save a *real* file with the data payload and then upload it to the
remote service (think about tmpfile()). Ugly and slow of course. So I
for the time being, I have given up in my efforts to use Curl, it just
doesn't behave well, maybe I will try again in a later version of php
or someone will give me another hint.

-- Zlib compression
It's a great idea and there was a discussion in Hessian mail list some
time ago about how to incorporate it into the protocol. The C++
version of Hessian, HessianCPP, uses it but it does recognize gzip
compressed data by looking a the first 2 or 3 of the data itself, not
by a protocol standard. People has proposed several options to
incorporate in into the protocol, or keep it simple and use HTTP
headers to see if data needs to be decompressed. I am working in the
new version of HessianPHP and I am planning to incorporate this
feature as soon as possible.
Check out Hessian's original mailing list for a deeper discussion:
http://www.caucho.com/support/hessian-interest/