Okay, so multiple write locations is not presently possible. Assuming I limit my application to one writer, what will happen in these two scenarios? Are locks protecting the data in these events?
1) What happens when a reader is active and a write request is made? 2) What happens to a reader if the writer is active? ----- Original Message ----- From: <[EMAIL PROTECTED]> To: <[EMAIL PROTECTED]> Sent: Wednesday, April 14, 2004 5:00 AM Subject: Metakit Digest, Vol 6, Issue 4 > Send Metakit mailing list submissions to > [EMAIL PROTECTED] > > To subscribe or unsubscribe via the World Wide Web, visit > http://www.equi4.com/mailman/listinfo/metakit > or, via email, send a message with subject or body 'help' to > [EMAIL PROTECTED] > > You can reach the person managing the list at > [EMAIL PROTECTED] > > When replying, please edit your Subject line so it is more specific > than "Re: Contents of Metakit digest..." > > > Today's Topics: > > 1. Re: Questions (Brian Kelley) > > > ---------------------------------------------------------------------- > > Message: 1 > Date: Tue, 13 Apr 2004 10:09:44 -0400 > From: "Brian Kelley" <[EMAIL PROTECTED]> > Subject: Re: [Metakit] Questions > To: Ray Mosley <[EMAIL PROTECTED]>, [EMAIL PROTECTED] > Message-ID: <[EMAIL PROTECTED]> > Content-Type: text/plain; charset=ISO-8859-1; format=flowed > > There is a pretty good description of the extend/aside mode here (in > python, but it should be easily transferable) > http://aspn.activestate.com/ASPN/Cookbook/Python/Recipe/252494 > > You will still have the problem with one writer and simultaneous > readers though. > > metaserver controls this by utilizing a metakit proxy that marshals > all read/view/write operations through a central server. In practice > it has been really fast although it does have a problem with > transactions and rollbacks are kind of meaningless since any data no > commited by any user will be rolled back. It is in python only > though, Jean-Claude and I are working on a standaridized mkrpc but > this is a little ways off. Again the prototype is in python. > > We are working on methods around this, in many ways the extend/aside > mode is a good start except that it needs to be extended to support > not just one aside file but many aside files (one for each user) but > this becomes tricky when you consider the need for table locking or > row locking so users don't clobber each other's data. > > So in short, multiple readers can access the data using extend/aside > but you will have to marshall all data changes through the writer. In > practice, this works very well for me because my frequency of data > committing is small compared to the number of reads. > > I hope this helps. > > Brian Kelley > > > ------------------------------ > > _______________________________________________ > Metakit mailing list - [EMAIL PROTECTED] > http://www.equi4.com/mailman/listinfo/metakit > > > End of Metakit Digest, Vol 6, Issue 4 > ************************************* > _____________________________________________ Metakit mailing list - [EMAIL PROTECTED] http://www.equi4.com/mailman/listinfo/metakit
