Re: load time regression: Re: commit: Change Record UUID's

Hi,

I can't claim to understand what's going on here, but how about using
the timer when the document is loaded (and/or saved?) to generate a
document-unique ID, and then appending a counter value to this for
change records?

Frank

On 23 Oct 2005, at 12:05, Tomas Frydrych wrote:
>> Hmm This is a very good point. Tomas why do we need such a complex
>> structure? In fact I think we need just a simple counter that
>> increments
>> with every change record created.
>
> Incremental counter is not good enough, since the puprpose of the ids
> is to differentiate between change records generated from separate
> instances of AW, and as such they need to be unique in time and space.
>
> As said elsewhere in this thread, the problem on win32 is to do with
> the 15ms granularity of the time API, but we cannot avoid it, because
> the collaborative stuff will require accurate timestamps. I understand
> there are ways in which a more accurate time can be got on win32, and
> if someone wants to investigate this to improve the performance of the
> UT_Win32UUID class, patches are always welcome.
Received on Mon Oct 24 09:07:38 2005