SSL compression

[nifoc]

OpenSSL 0.9.8+ supports SSL compression (i.e. compression of the entire SSL stream), which should increase the throughput and lead to "faster" download speeds. Some providers (like Hitnews) support SSL compression, so it would be nice for SAB to support it, too.

In Python 3.3 this is supported out of the box, but for 2.X it looks like you have to use pyOpenSLL.

[shypike]

Any data on whether it's actually worth it?
It would be a waste to spend a lot of time on a feature that has limited
support and may not bring all that much.

[sander]

I did a test with compression in scp with two different types of files:

A .nzb file, thus text:

Code: Select all

scp -C -v -c arcfour Wii-blabla.nzb   td6.xyz.net:~

debug1: compress outgoing: raw data 1146718, compressed 313825, factor 0.27
debug1: compress incoming: raw data 224, compressed 158, factor 0.71

So: compression from 1.1 MB to 0.3 MB ... great!

Test with a real article in a SAB download incomplete directory:

Code: Select all

scp -C -v -c arcfour SABnzbd_article_wEW3Xh  td6.xyz.net:~

debug1: compress outgoing: raw data 384368, compressed 381849, factor 0.99
debug1: compress incoming: raw data 178, compressed 139, factor 0.78

Conclusion: almost no compression (only -1%) of an actual article, so no advantage of using SSL-compression on usenet articles (which are already compressed).


EDIT: the " -c arcfour " is the type of encryption, and not needed for / related to compression, so it can be left out. The result is not different.

[shypike]

It's likely that Hitnews supports it for text groups, where it is useful.
yEnc-encoded RAR data has so much entropy, that compression is not feasible.
Like rarring a zip file is useless.
That's my non-expert explanation...

[nifoc]

I did a test with compression in scp […]

Should've tested this before I requested it …
I can basically confirm your numbers (768000 Byte uncompressed Vs. 766493 Byte compressed), so implementing this is pretty useless.