Re: [Tzdist] Test Server

[Paul Eggert <[e-mail]>]

Eliot Lear wrote:
>    * There are some semantics tied in with those names- it's not just
>      versioning.  I think that's asking for complexity, but maybe the
>      complexity is already there?

Yes and no.  The tz database has several configuration options, but each 
has a default and the plain IANA version 2014f would surely just use the 
defaults.  Downstream publishers such as Apple and Debian often change 
the defaults (or even the data), but they're in different namespaces.

This raises a problem I have with Figure 1 of 
draft-douglass-timezone-service-11: its model of Contributors and 
Publishers doesn't match my understanding of how things work.  In 
practice the tz database acts as a sort of clearinghouse or "root 
publisher": downstream publishers like Debian republish modified 
versions and occasionally contribute patches back, and that's how the 
tz-mediated universe stays reasonably consistent even though minor 
details differ.  Figure 1 implies that contributors and publishers 
maintain independent databases, which isn't really what's going on.