Nelson Bolyard wrote:
Most DBs have licenses incompatible with MPL. I think the mozilla/NSS developer community doesn't want NSS to depend on a DB that isn't as readily reusable and redistributable as the rest of NSS is under MPL.
Certainly. What do you or other NSS responsable have to say about Keith Bostic's message ? Or maybe you're discussing it in private ...
I think Gervase's recent answers in this thread speak for most of us.
There are also some technical requirements for a replacement DB. It should pass the ACID test.
(for anybody needing more details : http://www.archwing.com/technet/technet_ACID.html)
I don't believe the current db passes that test.
If you're referring to the DBM now compiled into NSS, you're right. It doesn't. That's the cause of this issue, of course.
> Many people were using mysql without the atomicity.
Do you mean that if several process are to access it, then it becomes an absolute requirement ?
Yes. Today, NSS uses PRLocks within a process to achieve the required atomicity and isolation (more or less) among threads within a process. Those characteristics are needed by NSS, and a multi-process solution must offer them.
Perhaps some work will need to be done in NSS to add begin/end transaction calls (not now present) once we find the suitable candidate.
Well if operations are not atomic now, why do we require them to be after ? Can't we handle some of the atomicity inside NSS ?
Again, today, the atomicity is not achieved through calls to DBM (which does not provide them), but through other means. When the DB provides them, the code may need to be changed to get this from the DB rather than from these other means. Bob said that code has been written. My *guess* is that it hasn't been thoroughly tested, but I'm a pessimist.
_______________________________________________ mozilla-crypto mailing list [EMAIL PROTECTED] http://mail.mozilla.org/listinfo/mozilla-crypto
