Re: [Exist-development] Read lock problem in WebDAV causes database deadlock (Potentially Serious)

[Adam R. <ad...@ex...>]

>> So what did you make of my proposals (August 8th) for changes to the
>> code, and can you comment on why we release the internal READ lock but
>> tell WebDAV that its locked - that seems dangerous to me, but there
>> are probably some webdav specific reasons?
>
> The internal locks should *never* be held across invocations. They
> have to be released before a transaction commits or you'll block other
> threads.

If the lock is released before the transaction is committed is it not
possible the thread could be preempted and the resource modified by
another thread and committed before this one is committed?

> The WebDAV (or user level) locks are for applications to implement
> their own locking scheme. It is up to the application or interface to
> enforce those.

But what happens if WebDAV or another applications, thinks that it has
locked something in eXist-db, but it has not and another API removes
this resources. Surely locking has to be end to end?

> Wolfgang
>



-- 
Adam Retter

eXist Developer
{ United Kingdom }
ad...@ex...
irc://irc.freenode.net/existdb