[cgl_discussion] Re: [OCF]draft 0.8 of the SAF Application Interface specification
lmb at suse.de
Mon Jan 13 03:45:55 PST 2003
Joe DiMartino <joe at osdl.org> said:
> The first two drafts and the most recent draft of the OCF event service
> use an opaque object that can be converted into a file descriptor. We
> DO assume that the underlying OS has a mechanism similar to poll/select.
I think - for the target audience of OCF - assuming basic POSIX compliance is
probably safe, maybe we could even assume suv3, but indeed not as readily.
I think _without_ making _some_ assumptions - like POSIX - the standard would
become an unwieldly complicated beast with three heads. Now, a clustering
standard is always going to be a complicated beast, but I think we can do with
2 heads ;-)
> I think it was understood, if not explicitly stated, that OCF would use
> the openDLM in its full glory, complete with lock conversion and async
> notification of blocked lock requests. At least that was my plan.
I think this has indeed been our plan; I'm in full agreement here.
Watering down the DLM makes it less useful; I am all for 'Keep it simple,
stupid!', but there is such a thing as _needed_ complexity. I think it is safe
to assume that any DLM which can't support a database and a CFS is not
powerful enough and will lead to people bringing their own DLM again - in
which case the point of standarizing an API is kinda lost.
At least the specification must allow the advanced features, even if they may
not be implemented everywhere - an application could then just bail out at
startup if it doesn't find the required features.
> Any attempt to water it down seems naive.
Lars Marowsky-Brée <lmb at suse.de>
SuSE Labs - Research & Development, SuSE Linux AG
"If anything can go wrong, it will." "Chance favors the prepared (mind)."
-- Capt. Edward A. Murphy -- Louis Pasteur
More information about the cgl_discussion