Re: ISSUE: Checkgroups control messages

On 25 Sep 2008, at 02:33, Charles Lindsey wrote:
> In <[email protected]> Thorfinn <[email protected] 
> > writes:
>> *nod* 64 bit is becoming fairly prevalent, and that should certainly
>> be enough.
>
>> That said - the DNS world is pretty used to using a 32 bit integer
>> circle for comparison.
>
>> Looking at the current USEPRO text... *ponder* Serial number MUST
>> increase.  That is definitely a problem - you certainly can run out  
>> of
>> 32 bit integer.  It's not all that likely, but it's certainly  
>> possible.
>
> Well, it has taken 30 years for a few heavily populated groups to come
> within sight of overflowing the 32-bit article numbers, and the  
> number of
> checkgroups messages sent is many orders of magnitude less than what  
> such
> groups have seen. So I think 32 bits is fine. People still regard 32  
> bits
> as the 'normal' length of an integer, even though modern hardware is  
> able
> to cope with 64bit, especially for addresses.


*nod* I'm pretty comfortable with no numerical limit on checkgroups  
serial number (i.e., stick with the current text).  There's no good  
reason to treat the serial number as an binary integer at all... it's  
perfectly orderable using trim leading zeros, then compare with string  
length followed by ascii ordering after that.  It's not exactly a  
comparison you have to do very often.

Offtopic:

NNTP article numbers, different story - there are, after all, rather a  
lot more of them on a busy news server, and they are potentially  
needed as index entries and such.

Sequence space arithmetic is definitely good if you're playing with a  
limited number space - the DNS folks had it right way back then. :-)

Meep,

  Thorf

-- 
<a href="http://tertius.net.au/~thorfinn/">[email protected]</a>
The world is coming to an end.  Please log off.  -- BSD fortune file