[Twisted-web] Nevow future

glyph at divmod.com glyph at divmod.com
Wed Apr 12 12:22:27 CDT 2006



On Wed, 12 Apr 2006 15:01:46 -0200, Manlio Perillo <manlio_perillo at libero.it> wrote:
>glyph at divmod.com ha scritto:

>But, as an example:
>
>wc channel\fastcgi.py
>  343  1078 11712 channel\fastcgi.py
>
>wc channel\http.py
>  868  2795 31479 channel\http.py
>
>
>So, http "support" is far more difficult to implement?.

The difference is actually smaller than that, since HTTP support is better commented and documented.  Also, the HTTP channel has a lot more features, such as client timeout support and pipelining.  The FastCGI support, as you mentioned, is incomplete, not supporting the (broken) multiplexed connections described in the protocol, or AUTHORIZER or FILTER roles.

Also, they both import a great deal of logic from twisted core and other areas of web2.  The difference is even smaller when you take the full picture into consideration.

I think what this difference boils down to is that a lot more time and effort have (rightly) been spent on the HTTP channel than the fastCGI channel; totally correct, robust support for both would likely be close to equivalent.

>> Yes.  Personally, I love lighttpd; I would really like it if, for
>> example, when Mantissa starts a web server, it could automatically put
>> static content in a front-end lighttpd proxy.  I should really file a
>> ticket for that...

>I'm not a lighttp developer, sorry ;-).

It is a mantissa ticket, not a lighttpd ticket, anyway :).

>>> it is possible to forward an HTTP request using Unix domain sockets?
>>> Is this supported with Apache and lighhtp?

>> Not as far as I know.  Everything sucks except Twisted ;-).

>Maybe a file ticket for lighttp is needed here too.

Probably.  It may be possible, I haven't checked.

>>>> Yes but _WHY_ don't you like the idea?

[...]

>Well, thanks.
>I got the point.

OK.  I won't belabor it any longer, then :).



More information about the Twisted-web mailing list