Re: [sip-clf] New CLF Syntax draft (text with index)

[Adam Roach <[e-mail]>]

Simon Perreault wrote:
> Adam Roach wrote, on 07/05/09 03:20 PM:
>> I considered doing it both ways, but decided that I preferred the slight
>> space impact of being able to get to any fixed field by looking at only
>> two numbers instead of all the preceding fields. I'm open to doing just
>> lengths if people are very sensitive to the size of each record, but
>> have a personal preference for what's in the draft currently.
>
> I'm sorry, I don't understand. Why would one need to look at all the preceding
> fields?
>
> Let's say you want to look at field 3. You look up the third pointer (p3), this
> gives you the start of the field. Then you look up the fourth pointer (p4), this
> give you the end of the field. If you want the length you compute p4 - p3.

Ah, yes! That's a good point. I had considered encoding *lengths* only, 
which would require adding all the preceding fields together (and I had 
mis-read your suggestion to mean lengths instead of pointers). You are, 
of course, correct -- encoding only pointers gets me what I need, while 
cutting the index size in half. Thanks! I'll try to get a revised draft 
out reflecting this change sometime soon.

/a