[zeromq-dev] Fwd: Exact matching on subscription topics
andrew at research.att.com
Thu Jan 19 02:53:42 CET 2012
he literally meant this thread; the subject line is above.
as for regex, be careful. regex performance is NOT wire speed EXCEPT
for the case of substring and in many case, multiple substrings
(the unix fgrep case). plus, relatively few folk are adept with complex regex expressions.
on the other hand, i don't get this use case.
can anyone share why we need this complexity of subscription stuff?
someone mentioned having tens of thousands of "topics" -- what is this about?
On Jan 18, 2012, at 6:22 PM, john skaller wrote:
> On 19/01/2012, at 12:00 PM, Martin Sustrik wrote:
>> However, few simple filtering algorithms (exact, prefix, regexp?) that can be performed at wire speed may be in scope.
>> The question here is how to implement the latter.
>> The important point is that the filtering algorithm should be distributed to every node in the message distribution tree. Thus, supplying a callback function at the subscriber won't do.
> Ah, I see. Ouch. now you want "persistence" for functional data so you can transport it.
> The issue then become how to encode it. Regex looks appealing because the encoding
> is a string.
>> I suggest reading the past posts in this thread. There's a disucssion of the algorithm that would allow for new filtering algorithms that wouldn't break the old intermediate nodes not aware of the new algorithm.
> Ok, you don't have a rough data and topic name so I can search the archives?
> john skaller
> skaller at users.sourceforge.net
> zeromq-dev mailing list
> zeromq-dev at lists.zeromq.org
Andrew Hume (best -> Telework) +1 623-551-2845
andrew at research.att.com (Work) +1 973-236-2014
AT&T Labs - Research; member of USENIX and LOPSA
-------------- next part --------------
An HTML attachment was scrubbed...
More information about the zeromq-dev