Re: [MORG] fuzzy search

[Timo Sirainen <tss at iki.fi>]

On Mar 25, 2009, at 12:33 AM, <Zoltan.Ordogh at nokia.com> wrote:

> "However, the result MUST NOT contain messages that don't have a  
> \Seen flag set or whose size is 5000 bytes or more."
>
> I do not understand why the size limit cannot be fuzzy. What happens  
> when you find a match that is 5016 bytes?  Instead of making it "not  
> found", it could get a lower score.
> Same for seen and unseen.

Originally I had been thinking about a bit different syntax where this  
text made much more sense. Then Alexey suggested a better way, and  
yes, I did start wondering if this made sense afterwards but decided  
to just post the draft as it is and see what other people thought  
about it.

I can think of one potential reason why it might not be good to allow  
it: If it was allowed, is it a MUST, SHOULD or MAY that 5016 bytes can  
find 5000 bytes message? And if so, what should be the limit when it's  
dropped out of the search results? Or should it always be there, just  
with a smaller score (minimum score)? If there is no recommended limit  
for servers, is there a recommandation to clients when they should put  
such queries inside the FUZZY rules?

I suppose another way to look at it would be to look at it from user's  
perspective. Anything inside FUZZY should be supplied by the user.  
User may want to look at "about" 5000 byte message, and server should  
prioritize the results accordingly and without restrictions to  
implementation. Same as with strings.

> I am not sure what to say about SORT - could yield funny results - I  
> would need to see it in action.

I'm not sure what you mean by this. SORT is the main reason why the  
whole relevancy even exists.

> BTW: are you using the right boilerplate?

I think so. It got changed somewhat recently and I think I'm using the  
same as e.g. STATUS-IN-LIST (but too lazy to check right now).