Re: [Uri-review] [Fwd: I-D Action:draft-melnikov-mailserver-uri-to-historic-00.txt]

[Alfred Hönes <[e-mail]>]

Mykyta,
thanks for your contribution.  A few comments inline below.

> Hello all,
>
> Some notes on discussed topic. At the moment
> there are some implementatiom of AFS mentioned
> below. One of them is in the Linux kernel starting
> from 2.6.10, but it is incomplete. There are some
> other implemenations. You can find some information
> on this topic here:
>
> http://en.wikipedia.org/wiki/Andrew_file_system

Yes, I'm well aware of implementations of AFS that reach far beyond
the original proprietary range.  However the question here is ...

a)  whether it makes sense to use a special URI scheme to access AFS
    related resources.

    All I could find is that due to the way AFS is used and the
    variability in the mount points et al., using an identifier that
    refers to the AFS server does not help in practice; 'file' URIs
    can be used on AFS clients for mounted AFS volumes since the
    mounted volumes are intended to behave transparently, like local
    file system volumes -- same as for NFS, e.g.

b)  in case the above is answered afffirmative, whether there is an
    accepted specification for afs: URIs, with interoperable usage.

AFAICS, neither the page you point to above nor the references
quoted in the recent RFC 5864 (on DNS SRV Records for AFS) contain
indications that would lead me to evidence that the above questions
have affirmative answers.


> As for tns3720 it is a telnet scheme for access to
> IBM 3720 terminal. As this terminal were started to
> be produced in 1972, so now it is OK to
> move it to 'Historic'. This service is not
> being used any more.

Well, AFAIK, 3270 emulation is still in some limited use in commercial
legacy mixed-platform environments, but I've never heard of practical
usage of the 'tn3270' URI scheme in that context -- all such business
applications are likely to operate based on tightly controlled local
configuration of the access points and not driven by some kind of
browser software operating on 'tn3270' URIs.

There's a single RFC (RFC 3049) that contains a few embedded examples
of 'tn3270' URIs, but it remains rather vague and does not give a
reference for the definition of such scheme either.

So I'm also in support of declaring that provisional URI scheme
Historic.


> As for AFS, I think it shouldn't be moved to
> 'Historic' - there are some implementation
> and it remains been used somewhere.

See my remarks above.
Do you have more information on definition and use of 'afs' URIs?
The thread I had pointed to yesterday seemed to indicate that active
AFS users didn't like the idea at all (already 17 years ago!).


> As for 'mailserver', it really needs to be moves
> to 'historic' as there are really no specifications of use.

Indeed.

The IETF process is very fast in this case.  IETF LC already has been
announced today for draft-melnikov-mailserver-uri-to-historic-00.

So I conclude that we both are in support of that draft. :-)


Kind regards,
  Alfred.


>
> Thank you for your time.
>
> All the best,
> Mykyta Yevstifeyev
>
> 17.11.2010 22:49, Alfred Hönes wrote:
>> My apologies for cross-posting, but this should perhaps reach a
>> broader audience.   I suggest to send feedback to the uri-review
>> at ietf.org mailing list, because Alexey sent his I-D announcement
>> there as well.
>>
>>
>> At Wed, 17 Nov 2010 17:26:32 +0000 , Alexey Melnikov wrote:
>>
>>> FYI. I've asked on Apps-discuss about deprecating the mailserver:
>>> URIs and I've heard several voices in favor of doing that.
>>>
>>> I will be asking Peter Saint-Andre to sponsor this document.
>>>
>>> ...
>>>
>>>     Filename        : draft-melnikov-mailserver-uri-to-historic-00.txt
>>> ...
>>
>> There are two more, apparently natively orphaned URI schemes listed
>> as 'provisional' in the IANA registry, both mentioned in RFC 1630
>> for the first time, and later again in RFC 1738, but never formally
>> registered and, AFAICS, still lacking any publicly available
>> specification:
>>
>>     a)  tn3270
>>
>>     b)  afs
>>
>>
>> A quick Google search does not reveal any hints on definitions and/or
>> actual use.  Some W3C web pages consider these as "reserved" names.
>>
>> Apparently none of the various RFCs on Telnet extensions for TN3270
>> and the somehow related MIB module specifications mention the 'tn3270'
>> URI scheme.
>> Even stronger, for 'afs', archived mail (from 1993) seems to confirm
>> that 'afs' as a URI scheme was a very poor idea from its beginning
>> (see "afs: considered harmful",
>> <http://wwwwbs.cs.tu-berlin.de/html/urls/0148.html>  ).
>>
>> I suggest to consider deprecating these reserved URI scheme names
>> as well, in a single draft together with the 'mailserver' scheme
>> (to save work in the process).
>>
>> Thereby, the registrations would be moved to "Historic", which,
>> according to the definition of that RFC status, does not immediately
>> invalidate existing use cases (if any), but serves to clarify that
>> new implementations should not try to support the subject.
>> The scheme names would remain in the registry, thus keeping the names
>> "reserved", i.e. out of availability for incompatible new use.
>>
>> So there are a few questions, for both cases:
>>
>> 1)  Is anybody aware of a specification for the above provisional
>>      URI schemes ?
>>
>> 2)  If "yes" for 1), are there interoperable implementations?
>>
>> 3)  Are these URI schemes in actual use anywhere
>>      (even if only intended for interpretation by brainware) ?
>>
>> 4)  Do the answers to the above questions support deprecation?
>>
>>
>> Kind regards,
>>    Alfred.