-----BEGIN PGP SIGNED MESSAGE-----
Hash: SHA1

Nicolas Cannasse a écrit :
> KaalH! a écrit :
>> -----BEGIN PGP SIGNED MESSAGE-----
>> Hash: SHA1
>>
>> Nicolas Cannasse a écrit :
>>>>> Another possibility would be to have an abstract protocol API that is
>>>>> implemented differently in mod_tora.c (or lighty_tora.c).
>>>>>
>>>>> Nicolas
>>>>>
>>>>>
>>>>
>>>> This is a tora proxy backend implementation for
>>>> lighttpd 1.5.0 >= r1992 :)
>>>>
>>>> lighttpd proxy layer allow load balancing & failover for free, that's
>>>> why I've use it instead of a standalone module.
>>
>> I've optimized the lighttpd patch, I've remove some duplicate buffers
>> and now directly use temp files, updated version attached.
>>
>> It's still parse all multipart data from temp files to find boundaries
>> positions before starting to send anything to tora, which can be quite
>> long, so timeouts can occurs with (very) large uploads.
>>
>> You set the tora socket timeout to 3 seconds, it is possible to increase
>> it by default without side effects ? or having a command line option for
>> it ?
> 
> I'm not a big fan of inscreasing the timeout, since it shouldn't depend
> on that for handling large files. Is it not possible instead to stream
> the data while parsing it like mod_tora for apache does ?

With other proxy backends (fastcgi,ajp13,http,scgi), raw data is sent as
is or splitted in fixed length data packets, final decoding is made by
the proxied server.

Adding smthg like CMultiBoundary & CMultiData to the tora protocol and
decode data in the server would be better for lighttpd.


I've may found a bug with getMultipart() during my tests, data seems to
be kept in memory, consecutive requests increase tora memory consumption.

> 
>>> Thanks, I'll have a look at integrating that in mod_tora. I'll maybe
>>> refactor a big chunk of protocol.* in order to separate the apache parts
>>> from the other ones.
>>
>> Great, tell me when done then I'll update the lighttpd patch.
>>
>> Do you think we should create a page on the haxe wiki for this patch
>> now, or is it better to wait for protocol.* refactoring ?
> 
> I'll try to refactor soon, so better wait a little more.

ok

Kaalh

> 
> Best,
> Nicolas
> 
> 

-----BEGIN PGP SIGNATURE-----
Version: GnuPG v1.4.6 (GNU/Linux)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org

iD8DBQFI6henWRw620CDc80RAuUJAJ9Zl04UryjgDBveZPr41+5sfw3/ZACfdqit
sC5wbraKRDfgOYVNmd4j11k=
=dy62
-----END PGP SIGNATURE-----

-- 
Neko : One VM to run them all
(http://nekovm.org)

Reply via email to