Hey, Eirikur

2006/7/21, Eirikur Hrafnsson <[EMAIL PROTECTED]>:
what problems could I run into if I use sequential-mode = OFF ?

This completely disables Slide's internal locking. It might cause some
weired concurrency errors as Slide's internal cache only supports an
isolation level similar to read committed. If you set your databases
isolation level to read committed as well this should be more
consistent. However, this isolation level might not be appropritate to
what you do.

Another secure solution would be to switch off caching and set
database isolation to serializable. The performance penalty is
disastrous, though.

We're running webapplications on Slide with various databases and today we
are using "WRITE" mode but I find it bothersome that a whole folder is
locked for read access when a long write is happening to one of its files.

Does fine-grained work better?

Fine grain locking only locks those folders that are actually
concerned by an action. This may lead to some form of - very hard to
explain - distributed deadlock, though. Depending on your database and
then again isolation level.

An alternative is the option to retry upon conflict. This comes at a
general performance penalty, but may be a good option.

More details at

http://wiki.apache.org/jakarta-slide/ConcurrencyIssues

For example a webapp we created automatically gets RSS feeds and stores them
in Slide every 20 minutes. The frontpage of the webapp then tries to read
these files via an Ajax RSS reader and has to wait until the writes are
finished. I would much rather (if versions are enabled) that while the PUT
is going on the GET would get the last version available but otherwise wait
until the PUT is done and only that resource gets locked up for the moment.
Is that possible today?

Might work if you upload to a diffrent file name and later move it to
the desired one. Move generally should be be faster than put.

And the last question...when will the next stable version of Slide be
released?

Don't ask me. I have no idea.

Cheers

Oliver

---------------------------------------------------------------------
To unsubscribe, e-mail: [EMAIL PROTECTED]
For additional commands, e-mail: [EMAIL PROTECTED]

Reply via email to