[Date Prev][Date Next] [Chronological] [Thread] [Top]

Re: slapd internal API change




On Saturday, March 29, 2003, at 11:36 PM, Howard Chu wrote:
back-perl - looking at Eric Stokes' patches still

Eric's patch does two things: add connection id to parameters passed to each perl function, and split searching into multiple functions to avoid deeply loading the perl stack for pathologically large searches.


The original patch is based on the old abi, so we'd need to change to use Operation instead of Connection.

But if we need to export per-connection data to perl, do we want to use a different method? Our stuff only needs an id to keep separate connections separate, but I could see potential uses for the other per-connection goodies in those structs.

If we do want to publish additional data to perlsubs, it seems like a bad idea to convert everything to perl data structures for every call. Does it make sense to instead provide a function (or functions) that the perl-side can call to get this data? Since perl is protected by a mutex already this should be relatively straightforward to implement.

Leif, how does this mesh with the connection_init / connection_destroy stuff you presented in San Francisco?

Or, if no one has any use for that we can just use the connid stuff...

Matthew Backes
lucca@csun.edu