RE: searching multiple backends

["Howard Chu" <hyc@highlandsun.com>]

> "Howard Chu" <hyc@highlandsun.com> writes:
>
> > Your understanding of directories is flawed. This is like
> saying the DSA for
> > c=US has to know everything, all in one database. It is
> analagous to saying
> > that a single root DNS server has to have total knowledge of every
> > registered domain. This is completely contrary to the
> distributed nature of
> > large-scale information systems.
>
> Do you really want me to answer to that?
>
> OK.  I know zilch about directories.  I don't know what they are used
> for nor how are they supposed to behave.
...
> Since your slapd server cannot locate the baseObject 'and thus has not
> searched any entries', it MUST return a referral resultCode.  That's
> what the -devel version of slapd MUST do.  If it does something else,
> let us know.
>
> If you, however, think it should behave differently, please tell us
> so, but don't forget to quote chapter and verse.
>
You stated that the directory layout I described was bogus. If you read
ITU-T Rec. X.501, section 19.3 "Directory Distribution Model":

	A master DSA's administrator may delegate administrative authority
	for any immediate subordinates of any entry held locally to another
	master DSA.

Figure 13 in this section further illustrates the point; I won't copy the
picture here but you can go find it and see.

> > > But I get your point.  I don't know the answer.  I don't even know
> > > what it does when a naming context expands several backends.  I
> > > think it should traverse backends transparently.
> >
> > Exactly.
>
> You are misreading me.  What I said is that namingContexts may span
> several backends of the same server.
>
> Julio

I don't believe I misread you at all. I gave a perfectly good example of
this exact situation, 3 backends spanning a particular naming context. And I
agree with you that the slapd should traverse these backends transparently.
But it doesn't do it today.
  -- Howard