Re: More History comments

Hi Rob,

Kevin Smathers wrote:

>> Hi Kevin,
>>
>> Your attachment shows the logical conclusion of following the logic I 
>> was trying to describe; that basically, if you change anything in the 
>> archive, since things in the archive are related changing one thing 
>> changes the situation that everything related to that changes.  This 
>> could prove to be a scalability issue; if every minor change in the 
>> archive results in a lot of things having a different situation, then 
>> the history system will rapidly become an enormous corpus of data.  
>> Maybe that's OK.  My point is that as far as I'm aware this vital 
>> issue hasn't been looked at even vaguely in the original or current 
>> work of the History system.
>>
>> Robert Tansley / Hewlett-Packard Laboratories / (+1) 617 551 7624
>>  
>>
> Ah, my apologies for being dense.
>
> I'm not sure that the corpus really grows that rapidly though as it is 
> only the nodes that have to be instantiated multiple times.  The 
> content pointed to by those nodes can shared between all of the nodes 
> that have that content (think C++ smart pointers, e.g: string class).  
> The overhead of the structure should be manageably small when compared 
> with the size of the data being pointed to.
>
> I do think that it is worthwhile (for size management, but also for 
> navigability) to avoid automatic archive of changes.  Instead archived 
> history should show 'relevant' changes, where relevance is determined 
> under application control.  For Simile that might mean only writing to 
> the archive once when a workflow has completed entry of a new 
> document, rather than seperately for each step in the workflow. 

Another alternative occurred to me on the commute home.  If the version 
changes are going to flow through an unreasonably large number of 
dependent nodes then one can use a CVS like approach, and bind a date 
(or name tags) to each node.  Then to view the graph as of a date you 
would find the most recent change that is not later than the date being 
requested.  This should work even if the graph looks like the one that 
you drew.


-- 
========================================================
   Kevin Smathers                kevin.smathers@hp.com    
   Hewlett-Packard               kevin@ank.com            
   Palo Alto Research Lab                                 
   1501 Page Mill Rd.            650-857-4477 work        
   M/S 1135                      650-852-8186 fax         
   Palo Alto, CA 94304           510-247-1031 home        
========================================================
use "Standard::Disclaimer";
carp("This message was printed on 100% recycled bits.");

Received on Friday, 23 May 2003 12:11:34 UTC