[Xcb] Query: is Xcb zeroing out unused fields in requests?

[Jim Gettys]

The Nomachine NX folks report that zeroing the datastream helps
significantly with data compression.

Whether this is correct or not, I don't know; adding data to the
discussion would be good (zeroing the bytes, and comparing the
relative compression using ssh -C, for example).

See:

http://www.nomachine.com/documentation/html/NX-XProtocolCompression.html

		Regards,
			- Jim


On Mon, 2004-11-08 at 11:41 -0800, Jamey Sharp wrote:
> On Sun, Nov 07, 2004 at 07:47:57PM -0500, Jim Gettys wrote:
> > If not, it should: the data stream will compress better...
> 
> It zeroes up to three padding bytes at the end of each part of the
> request (the fixed-length part, and any variable-length lists on the
> end). It has to, or incorrect behavior results for some requests. (I
> think QueryTextExtents is one.)
> 
> Do you have measurements that support this claim? I'm having a little
> trouble believing there's much impact.
> 
> I suppose we can memset the fixed-length request buffer to 0 on the
> stack before filling it in; alternately, we can generate code to zero
> explicitly-specified padding bytes. Both solutions bother me a little.
> _______________________________________________
> xcb mailing list
> xcb at freedesktop.org
> http://freedesktop.org/mailman/listinfo/xcb