[NNTP] NNTP URI draft

[Kai Henningsen]

chl at clerew.man.ac.uk (Charles Lindsey)  wrote on 14.03.05 in <IDC9u1.G8I at clerew.man.ac.uk>:

> In <9SfEP55Xw-B at khms.westfalen.de> kaih at khms.westfalen.de (Kai Henningsen)
> writes:
>
>
> >I think this points out a generic problem with the draft. It's not precise
> >enough about what it all means.
>
> >What the text SHOULD do, in my opinion. is, for every possible URI,
> >specify what sequence of NNTP commands it is supposed to cause, and what
> >the result is supposed to be, as much as possible. (Don't the http and ftp
> >schemes do that?)
>
> No, you can't do that, because it is not obligatory (for the news scheme
> at least) to use NNTP (though that will be the case 99% of the time). For
> example, when used with the local server, it might access the active file
> (whatever) directly. Moreover, even with an NNTP connection, it might
> choose to access the newsgroups file or the overview rather than the
> active file.

So? Say that if this is done via NNTP, this is how it's done; if it is  
done some other way, the result is supposed to be the same.

> >So, for example, news:bla at foo can be explained as causing ARTICLE blaa at foo
> >and returning the text returned by that command, if any, or an error
> >indication if that command returns one.
>
> That is fine for those forms of the URI that return a single article. But
> it is not so clear what is to happen to those forms that simply connect
> you to the server and leave it to you what to do next. If it connects you
> to a list of groups, is it then suppose to select one to read, or is it
> supposed to invite you to select one to subscribe to, or will it depend
> whether you are already subscribed?

Well, that certainly *should* be specified. Otherwise, it's useless.

>These are all things to be determined
> by the local implementation, depending on what sort of news reader it
> chooses to connect you to. The standard really shouldn't be caring about
> that.

No, the standard *SHOULD* specify that. A URI without specified semantics  
is a very bad idea.

> >Done like that, a lot of the semantic can be punted to the NNTP standard -
> >you do not need to care with what is a valid message id, or what happens
> >when there is no article for that message id, because the NNTP standard
> >already does - you just create inputs to NNTP and report outputs.
>
> I don't think there is much semantic to be specified, and I think my
> present text already goes as far as is justified.

I seriously disagree.

MfG Kai