[SPEC] Backwards-compatible metadata in Gemini

nothien at uber.space nothien at uber.space
Tue Feb 23 21:12:37 GMT 2021


Oliver Simmons <oliversimmo at gmail.com> wrote:
> Metadata is NOT for presentation.

By that, I mean things like the favicon and color proposals, which were
also bunched under the 'metadata' umbrella category/name.

> Myself I've more been thinking of the how and not the why with this,
> and I apologise for that.  Most of the examples given here are pretty
> pointless for clients, nobody has suggested any uses for metadata
> other than to replace the favicon thing, author, and the "when" of the
> document.  The only major uses I can think of is for orbits
> (webrings), and it would help search engines.  It's pretty pointless.

I agree, it's not useful, but I wanted to cover a few of the cases you
mentioned.  I disagree with the favicon feature as a whole, so I'm not
going to touch that.  I don't know how webrings are organized on Gemini,
but I'm not sure how the author or date would matter there.  From my
understanding, these things typically link to the whole gemini capsules,
which should already provide author information, and I don't think date
info is relevant here.  We already have the gmisub companion spec for
describing gemlog posts and similar dated content.  Search engines
typically date content based on when they first indexed it, so that's
not an issue either.

> > Read the spec, read the FAQ, read the companion documents, read the
> > mailing list, and /understand/ the spirit of Gemini.
> 
> I would hope everyone here has.

Unfortunately, it doesn't seem to be the case.

> All existing software would still work. This is only an addition.
> About the freezing, I got the impression it was only the protocol that
> was frozen.
> 
> ...
> 
> On Tue, 23 Feb 2021 at 19:23, Julien Blanchard <julien at typed-hole.org> wrote:
> >
> > We should always consider underpowered devices, screen readers,
> > people with a slow internet connection. I know TLS is already a
> > burden for some of these cases so let’s not add more data to
> > transfer.
> 
> I don't see how adding a few lines would cause much of an issue?

The issue is not that it would break pre-existing software, but that it
would make it less appealing to use, compared to new software which does
implement the new additions.  This applies to clients and servers and
everything in between.  That's the 'rift' I was talking about in my
original mail.

~aravk | ~nothien


More information about the Gemini mailing list