Re: [GNUnet-developers] End-user wishlist

[Tom Barnes-Lawrence]

Hi, me again...
On Wed, Jun 25, 2003 at 06:13:55PM -0500, Christian Grothoff wrote:
<snip>
> 
> I agree with that "afs" should be part of the URI. I would think of "afs" as 
> the authority that is responsible for the data. For the non-hierarchical hash 
> codes, I would suggest to use "." as a separator. That would yield:
> 
> gnunet://afs/HASH1.HASH2.CRC.SIZE
<snip>
> gnunet://afs/NAMESPACE/HASH1.HASH2.CRC.SIZE
> 
> Any further comments?

 As an alternative to using '|' I was thinking of '-' as a separator myself,
but '.' sounds just as good. The only disadvantage of '.' is if somebody
quotes the URL at the end of a (written!!) sentence. But that's not exactly
a fatal disadvantage, and I can't think of any advantage '-' would have.
Meh.

One thing I'd like to say, though, is that I remember the way Freenet did
their URIs. They came out with a URI scheme, like we are here, but then
as the network was primarily accessed through the http proxy, they used
http: URIs for that...
Except they decided they should use the freenet: URIs as PART of the http:
URI, like

  http://localhost:8080/freenet:address@hidden

Which Konqueror (the version I had at the time, at least) consistently
interpreted as

  http://veryverylongusername:address@hidden

because of the special meaning of : and @ in http URIs.

Now, the gnunet: URI being described doesn't sound like it would actually
have such problems, but the point is, that URIs should be used as URIs
(if you see what I mean...) and if you're actually accessing the resource
through a proxy with http, it should just pass the relevant arguments in
the way http usually accepts them.
But yeah, I'd much rather GNUnet had its own method of browser  access
(of some sort) that bypasses http, and (like, er, anonymous1 says) html,
in which case our own URI type would presumably be a requirement.

Tomble