On fre, 2008-05-30 at 09:59 +0200, Kinkie wrote: > Regarding WHAT to get in there, at this time IMO plaintext is fine. > Any other format would REQUIRE the intermediation of some software to > be made easily readable, and in my opinion that would be bad. OTOH, > adding an XML-based markup option might be useful for automated tools > if it's not hard to implement.
An interesting task would be to replace the SNMP agent in Squid with something more structured (and maintainable), and providing both SNMP and XML access to that data.. and add mib entries for the few interesting things we currently only have in cachemgr.. Adding an XML version of the current cache_object:// objects is not an easy task as it's all manual printf style formatting with no structure behind the data. Well, to be honest that XML formatter may just as well be a frontend on the SNMP agent, either in-process or as a separate service. I don't know why so many consider SNMP inefficient for statistics collection. It's a single UDP exchange each collection interval, collecting just the data set you ask for. And also a well understood protocol by most automata systems that may need to access such data.. The cache_object:// interface is really primarily meant to be used interactively by humans, and should not be considered a stable interface beyond the simple text formatting format that allows cachemgr.cgi to render it reasonably in tables and lists.. Regards Henrik
signature.asc
Description: This is a digitally signed message part
