[OGo-Developer] Re: additional field for synchronizatin

[Helge Hess]

On 24.08.2007, at 21:55, Wolfgang Sourdeau wrote:
> The chances that 2 users modify the same event (for exmaple, one  
> change the start hour and the other deletes it) are really close to  
> 0.0000001%.

The question is not how often a conflict happens per day or sync but  
per 5-years and not for 2 users but for 1000+ users (in the SOGo case  
many more) which synchronize their 1000+ events when they arrive in  
the office at 8:30.

Now if you are happy you might never encounter a conflict, but keep  
in mind that PostgreSQL can easily perform 10.000 updates per second  
on modern hardware so one second of granularity *is* quite broad.  
Then consider that concurrent changes to a shared item are much more  
likely than changes to some private, old or new item.

This keeps me concerned (but not overly, especially if the users are  
told about the potential issue). And its really only an issue for  
large deployments.

> So I would not worry about that at all.

As mentioned before I think its OK no to worry about it now. But  
maybe doing it transactionally safe isn't that much harder either?
If its too hard to bother, it fine for me.

Summary: I'm not exactly pleased about an implementation which relies  
on timestamps instead of exact states. But I'm even less pleased  
about not having an implementation at all :-)

So, no need to argue any further. I'm fine if you contribute an  
implementation doing it the way you outlined!

Thanks,
   Helge