groups with numeric name parts

groups with numeric name parts

[Bruce Stephens]

I note that son of rfc1036 rules out newsgroups whose first component begins 
with a digit (e.g., 3dfx.foo would be out, although foo.3dfx would be fine), 
but I don't see anything in 1036 itself that does.  (Indeed, I don't see any 
comment about names at all, and perhaps that's one reason for wanting 
something to be written down!)

I can see an implementation problem with name parts which are wholly numeric, 
and although I can see ways around that I really don't care about groups like 
that.  What are the arguments against turning a blind eye to things like 
3blind.mice?

Re: groups with numeric name parts

[Karl Kleinpaste]

Stainless Steel Rat <ratinox@peorth.gweep.net> writes:
> Early Arpanet/Internet standards describe how to do something, not how not
> to do something.  rfc1036 reserves numerics for articles, therefore they
> should not be used for other things.

Call me crazy, but you seem to have a distinctly peculiar incantation
of RFC1036.  In mine:

[1008] [20:39:11] pocari-sweat:rfc> grep -in numer rfc1036.txt
[1009] [20:39:17] pocari-sweat:rfc> grep -in numbe rfc1036.txt
288:    sequence number for messages submitted to the network, or a short
485:    reasonable number of backwards references.
578:    This contains a count of the number of lines in the body of the
585:    names and message numbers.  These are the newsgroups listed in the
586:    "Newsgroups" line and the corresponding message numbers from the
603:    the "Xref" line shows that the message is message number 461 in the
604:    newsgroup news.lists, and message number 6378 in the newsgroup
929:    Since news messages are usually short, and since a large number of
943:    followed by a message containing the given number of bytes.  (The
[1010] [20:39:24] pocari-sweat:rfc> grep -in reser rfc1036.txt
[1011] [20:39:30] pocari-sweat:rfc> 

Odd, no references to "reservations," or "numerics," anywhere.  In the
copy I've got, anyway -- maybe yours is different.

These "number" items in RFC1036 refer to...
288:     message-id
485:     references inclusion count
578:     lines header
585/586: xref header
603/604: example of xref
929/943: batching

Roughly translated, RFC1036 doesn't have the faintest idea about
proper choice of newsgroup names.

It helps if one first reads RFCs, before commenting upon them.

While I'm in an RFCoid rant, "early arpanet/internet standards" were
_/perfectly/_ happy to bitch and moan about an arbitrarily large set of
things that ought _not_ to be done.  How many dozen examples shall I
pull out, if you feel the point needs to be made explicitly?  Let's
start with the basics: "Be liberal in what you accept, and
conservative in what you send."  (First explicit reference: RFC1043,
merely making express note of a theme that had already been in place
for years.)  Now start defining the concept of "conservative" and what
it means one should _not_ do.  See RFCs 1122/1123 for the whole
MUST/MAY/MAY NOT/MUST NOT continuum; less than 2 years separates 1036
and 1123.

Theorem: Usenet systems send (have always sent; have always been able
to send) newsgroup names containing leading digits, and in fact have
always coped reasonably well with non-top-level name components
consisting entirely of digits (viz. alt.2600).
Proof:   Left as an archival exercise to those readers who weren't
actually around in the days of A News.

(Yes, I was.)

Given that netnews systems send such names (either in NNTP, or [as
they always have] by reading local active file and spool), all
newsreaders had damn well better be prepared to accept them liberally.

--karl

Re: groups with numeric name parts

[Sudish Joseph]

Stainless Steel Rat writes:
> should not be used for other things.  It does not explicitly state that
> numerics cannot be used for groups because the people who wrote the draft
> figured those implementing it would have a clue and not do something
> rock-stupid.

Or because they were far-sighted enough to realize that a protocol
should not be specified in terms of the limitations of the on-disk
file layout of one particular implementation?  

Anyways, in this particular case, even INN has no problem with
3dfx.*. 

-Sudish

Re: groups with numeric name parts

[Russ Allbery]

Karl Kleinpaste <karl@jprc.com> writes:

> Theorem: Usenet systems send (have always sent; have always been able to
> send) newsgroup names containing leading digits, and in fact have always
> coped reasonably well with non-top-level name components consisting
> entirely of digits (viz. alt.2600).
> Proof:  Left as an archival exercise to those readers who weren't
> actually around in the days of A News.

Counterproof:  Create a group named "alt" on your INN news server.  Type:

        ctlinnd renumber alt

Take a look at what it thinks the high number of the group is.  The last
time I tried this, it thought the high numbered article was 2600.  Guess
why.

alt.2600 has worked fine because there is no widely-propagated and widely
used group called "alt" to get hit by it.

-- 
Russ Allbery (rra@stanford.edu)         <URL:http://www.eyrie.org/~eagle/>

Re: groups with numeric name parts

[Karl Kleinpaste]

Russ Allbery <rra@stanford.edu> writes:
>         ctlinnd renumber alt
> Take a look at what it thinks the high number of the group is.

Thus, the exception which proves the rule.  The real, day-to-day issue
in question was the presence of digits, and especially leading digits,
in group name components, not all-digits components, which I was
merely noting in an also-ran sort of way.