[standards-jig] JEP-115 Redundancy?

Jacek Konieczny jajcus at bnet.pl
Thu Sep 25 11:24:35 UTC 2003


On Thu, Sep 25, 2003 at 11:25:46AM +0200, Ralph Meijer wrote:
> I think I saw this come by earlier, but I am not sure. What about having an
> identifier for the current configuration of a client? I am thinking about
> an opaque number or string. This would cover whatever the client feels
> worth querying, and not only capabilities, but also configuration variables,
> whatever those might be.

But this way you may only cache information per user. So if you have
1000 users in your roster (online? :-) ) you will have to disco them 
all (at least once) even if all of them use the same software with the
same configuration. If the client doesn't have any persistent storage it
would have to query all those clients in each session again.

I think one of the goals of JEP-115 is to avoid queries which are known to
give the same result. But the way it is defined in JEP-115 is not very
secure.

My proposition was to use identifier which depends only of the features
supported (and maybe other discoable information) and it is impossible
(very hard) to find two sets of features with the same identifier
- identifier would be eg. MD5 sum of normalized feature list.

Greets,
	Jacek



More information about the Standards mailing list