So it's really xupdate..., unfortunately I couldn't find the sources for the xupdate implementation or better a working cvs or subversion address. Do you have any that works?
According to your comment it sounds like that any two xupdates can conflict no matter on which document of which collections it is ever executed. So if a person A1 is editing a document B1 in a collection C1 and a person A1 is editing B2 in C2 via xupdates at the same time, there is a chance that one or both files are "corrupt"? So my workaround to synchronize xupdates per collection only decreases the possibility of corrupt files but does not eliminate it? So the only possible solution without touching the sources of xupdate (which I couldn't find as mentioned above) would be to synchronize all xupdates over the whole database or not to use xupdates at all?!? Bjoern Eickvonder PS: By the way where and in what kind of form can/should I post XIndice patches that I developed like a solution for the "too many open files problem" that can occur on linux systems (if you have a large number of collections)? > -----Ursprüngliche Nachricht----- > Von: Todd Byrne [mailto:[EMAIL PROTECTED] > Gesendet: Donnerstag, 28. Juli 2005 20:10 > An: [email protected] > Betreff: Re: AW: AW: Performance > > > The xupdate implementation that Xindice uses isn't thread safe. The > xupdate code uses a global static variable to hold temporary nodes. > Their is also others cases where temporary elements can be left around > but I haven't been able to get the patch commited on the xupdate side. > > Todd > > Eickvonder Bjoern wrote: > > I use b4 and it does occur if multiple users are manipulating > different documents in a collection via xupdates, especially > those commands that append elements. In that case sometimes some > of the appended elements are missing or a document contains > elements that would normally belong to another document. I > tracked this issue a little bit and I think it is maybe somewhere > in the xupdate implementation, that is used, so it is not > entirely an xindice problem. > > > > Björn Eickvonder > > > > -----Ursprüngliche Nachricht----- > > Von: Vadim Gritsenko [mailto:[EMAIL PROTECTED] > > Gesendet: Donnerstag, 28. Juli 2005 16:33 > > An: [email protected] > > Betreff: Re: AW: Performance > > > > Eickvonder Bjoern wrote: > > > >>Moreover I had problems if multiple users are adding/writing > >>(different) files to a collection concurrently (some data was later on > >>corrupt or missing), so I had to synchronize the write access to each > >>collection. > > > > > > What xindice version did you use. I've not seen corruptions since b4. > > > > Vadim > > > > > > > > > > ____________ > > Virus checked by G DATA AntiVirusKit > > Version: AVK 15.0.6273 from 27.07.2005 > ____________ Virus checked by G DATA AntiVirusKit Version: AVK 15.0.6302 from 28.07.2005
