> > >However, I don't think this is unreasonable. There is no requirement
> > >that tools be able to parse URIs to extract meta-data.

Say who?

There is a requirement that repositories "work" (at some minimum level)
without metadata, especially since we aren't specifying metadata. Without a
parsable URI (or parsable URL) how do tools read a repository to do things
like "clean oldest nightly/snapshot, but leave all releases", "download
latest release" or even the basics "determine/display contents", "show basic
contents" (irrespective of version/type).

If we are shooting for humans, not tools, why bother changing what we have?
Humans can grok it all, we don't need any "standard". If we are proposing a
standard, there has to be a valid purpose for it -- and having a standard
that isn't structured for computer processing seems setting the bar
pointlessly low.

For me, the strongest argument for tooling (other purely than saving admins
effort) is download + verify (MD5/whatever). A human is far more likely NOT
to do it, they'll right click in their browser, save and use. I think we
ought have a structured repository so we can write administration tools, but
also timer save "download" plugins that do the requisite verifications and
keep our users secure.

Again, a specification that limits/breaks tooling is wasting of huge



Reply via email to