[Date Prev][Date Next]
Re: subtype indexing issues
On Fri, 25 Feb 2000, Kurt D. Zeilenga wrote:
> I'm working on restructing indexing... one issue needing resolution
> is what exactly to index in the face of subtyping.
> Should indexing be explicit:
> "index name" generates index for (name=foo) matching?
> hence requiring:
> "index cn" to generate index for (cn=foo) matching
> "index name;lang-En" to generate index for (name;lang-En=foo)
> And should one use supertype indexing to limit canditates for a subtype?
> That is, use cn and/or name;lang-De indices to limit cn;lang-De searches.
> I guess I don't have much problem with the latter. The server should
> attempt to limit candidates via whatever indices are available.
> I do have a bit of a problem with allowing the "index type" generate
> indices for subtypes, even where we limited the subtypes to those of
> same type with description options. The problem is that each index
> generated is work which may be undesirable.
> On alternative is to allow a syntax like:
> index cn ... (generate indices for just CN)
> index cn;lang-En (generate indices for just CN;lang-En)
> index cn;+ ... (generate indices for CN with options)
> index cn;* ... (generate indices for CN and CN with options)
> index cn;lang-* (generate index for CN with language tags)
> I am thinking that all indices for the same attribute type
> (with or without options) should be stored in single file.
> This implies that key to option(s) specific indexes need to
> include the option(s).