gnutls-devel
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: GPLv3 migration reminder


From: Marc Lehmann
Subject: Re: GPLv3 migration reminder
Date: Thu, 20 Dec 2007 03:39:36 +0100

On Wed, Dec 19, 2007 at 01:04:47PM +0100, Simon Josefsson <address@hidden> 
wrote:
> Right, that's why I was interested in doing some real TLS-related
> benchmarks...

Count me in :)

> Network traffic.  Currently DEFLATE (libz) is supported by TLS, and
> there is one patented alternative (LZS) specified.  We support LZO too,
> even though it isn't standardized.  I don't think LZO in TLS has any
> wide deployment, and in any case, there is nothing that says we can't
> support both LZO and LZF.  Doing so would make it easier to perform
> benchmarks.

I'd say supporting lzf AND lzo makes little sense, as its yet
another incompatible format (even if its autonegotiated, ensuring
interoperability), and your problem seems to have been licensing with lzo.

I'd recommend staying with liblzo (it is far more flexible) if you have
to support it anyways, but you can of course do what you want, I won't
complain if you find you want (my) liblzf in gnutls, I'd be honored :->

(I would also say that network bandwidth usually should take priority over
(reasonable) compression times, and liblzf is certainly on the speed side,
not on the data compression side, of the spectrum).

> Still, I guess that there are more higher priority projects in GnuTLS..

Yeah, but thats what the TODO list is for. In rxvt-unicode, for example,
whatever you put on the TODO list instead of tackling it immediately will
almost cetrainly stay there for years.

> but compress benchmarks seems like an interesting thing to do.

Go for it, and tell me any results, thank you :)

If you do them, the default is probably good, but you can look at lzfP.h for
configuration options, like:

   HLOG        - for memory usage, lower is less memory
   VERY_FAST   - enabled "medium mode", the default, which is good for binary,
                 somewhat bad for text
   ULTRA_FAST  - make it even faster, and quite a bit worse, ratio-wise.

One could even play with the hash function, which is critical, but the
default will certainly do.

-- 
                The choice of a       Deliantra, the free code+content MORPG
      -----==-     _GNU_              http://www.deliantra.net
      ----==-- _       generation
      ---==---(_)__  __ ____  __      Marc Lehmann
      --==---/ / _ \/ // /\ \/ /      address@hidden
      -=====/_/_//_/\_,_/ /_/\_\




reply via email to

[Prev in Thread] Current Thread [Next in Thread]