[Standards-JIG] Q about shorter tag names

Ulrich Staudinger us at activestocks.de
Sun Nov 20 12:37:20 UTC 2005


Ralph Meijer schrieb:

>On Sun, Nov 20, 2005 at 01:11:05PM +0100, Ulrich Staudinger wrote:
>  
>
>>So be it for now ... For future versions of XMPP (i.e. 1.1) the replacement of
>>the long tag names should be considered nonetheless.
>>
>>Right now i am in a work group where we calculate and test with 500k concurrent
>>connections in a jabber cluster. Those long tag names increase our traffic by
>>the factor 4-5 approximately. We, in our scenario with the current specs, would
>>produce about 6-10 Terrabyte of traffic per month, with compression propably
>>something like 2-7Tb, with shorter tag names approx. 1-2.
>>    
>>
>
>Depending on what will actually be inside the packets, I believe you
>will find that compression will yield far better results. If shorter tag
>names will make your traffic shrink by a factor 5 as you state, you can
>be assured that compressing regular XMPP (1.0) traffic will exceed this
>factor.
>
>I suggest you actually try to find a representative traffic exchange for
>your application, generated by some test application, and then rerun the
>test with JEP-0138 enabled. Please share your findings. If things would
>be as you claim, we can always look into changing the protocol.
>  
>
Thanks Ralph,

i will try to publish the results as good as possible.

Cheers,
Ulrich
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mail.jabber.org/pipermail/standards/attachments/20051120/23f44447/attachment.html>
-------------- next part --------------
A non-text attachment was scrubbed...
Name: us.vcf
Type: text/x-vcard
Size: 329 bytes
Desc: not available
URL: <http://mail.jabber.org/pipermail/standards/attachments/20051120/23f44447/attachment.vcf>


More information about the Standards mailing list