Re: [MORG] I-D Action:draft-ietf-morg-fuzzy-search-01.txt

[Barry Leiba <barryleiba.mailing.lists at gmail.com>]

>> UIDs, sequences, modification sequences: [[anchor6: TODO: There is
>>    probably not much point in matching these fuzzily.  Could it be a
>>    MUST NOT for clients to use them inside FUZZY?]]
>
> I don't really have any thoughts on this. Seems pointless, but then
> again maybe it might make something easier to allow them to be matched
> (exactly?)

I think the point is that the server decides what "fuzzy" means, and
that the decision may vary from search key to search key -- including
doing an exact match because it thinks that "fuzzy" isn't meaningful
for that particular key.  I would never want a client to be unable to
say something like this:
  SEARCH FUZZY (FROM "siranan" UID 123:217)
...even though the client could put the UID item outside the FUZZY
scope.  It's up to the server how to handle it; I don't want the
client to be trying to second-guess what the server will do.

Maybe a change to the text will help.  How about this?:

OLD
Fuzzy matching is not limited to just string matching.  All search
keys SHOULD be matched fuzzily, although what exactly that means for
different search keys is left up to implementations to decide.  Some
suggestions are given below.

NEW
Fuzzy matching is not limited to just string matching.  All search
keys SHOULD be matched fuzzily, although what exactly that means for
different search keys is left up to server implementations to decide
-- including deciding that fuzzy matching is meaningless for a
particular key, and falling back to exact matching).  Some suggestions
are given below.


And then maybe:
NEW
 UIDs, sequences, modification sequences: These are examples of keys
for which exact matching is probably what makes sense.  Alternatively,
a server might choose, for instance, to expand a UID range by 5% on
each side.

---
What do you think?

Barry