In message <d11357c2.942b%[email protected]>, Edward Lewis writes:
> I admit that this thread rolled off my screen before I noticed it.  I’m
> going to pick on a few things said here that extend backwards in the
> thread a bit.
>
> On 2/22/15, 15:53, "Mark Andrews" <[email protected]> wrote:
>
> >master files are ASCII files.
>
> Maybe they are supposed to be.  But what got my goat recently was trying
> to parse a zone file that had UTF-8 in it.  There are zone master files
> that are not purely ASCII in the wild.

There are files that people want to be master files with UTF8 in them
but they are not master files.

> Not saying that it is right.  Saying that they deserve to be recognized.
> I mean, there’s every reason that the file can be put to use.  I had
> considered extending the parser I had to do so, then decided for my use
> case the effort wasn’t worth it.
>
> >If you have a file that contains u-labels it is not as master file.
>
> Walks like a duck, talks like a duck, smells like a duck…but yes,
> according to some dusty old documents - written before u-labels were
> imagined [by the community producing the documents] - they aren’t.

And also that IDNA explicity does not require changes to nameservers.

> >It can be converted to a master file by processing the labels and
> >turning them into a-labels.
>
> Bingo!
>
> >Master files are supposed to be able to be processed by nameservers
> >that are not IDNA aware.
>
> There’s no reason that we can’t add an adjective to "master file.”  Then
> folks writing RFPs, contracts, work orders can specify what is best.  If I
> choose to run antiquated software I’ll ask for antiquated master file
> s. If I choose software that reflect the native language of my staff, I might
> want master files that have a different formatting.

I repeat.  They are not master files.  They a 'u-files' or whatever
name we decide to give them.  That said non-ascii doesn't mean that
the file contains UTF-8.  Lots of "master files" contain ISP-LATIN-FOO
or Windows-XXXXX.  Making any assumption about the character set
is plain fraught with danger.  IDN goes native -> utf8 (u-labels)
-> a-labels.

> Orthogonal to my desire to document this is the desire to see how far we
> can get away from assuming everyone has ASCII knowledge.  NOCs shouldn’t
> have to list a particular (human) language as a requirement.
>
> >If you use master files extensions such as $TTL, $DATE and $GENERATE
> >then you need to ensure that the nameserver can process them.  A
> >cannonical version of the master file contains none of these
> >extensions.
>
> I know of name servers that do not read master files at all, so no, you
> can’t “ensure."

Master files are a zone transfer method.  You only need a tool to
convert them to whatever you are using internally to meet RFC 1034
requirements.

i.e.  here is a master files for zone example.net.  Please load it
into your nameserver.

There is no requirement for a nameserver to read then directly though
many do.

If you "extend" the format you need to know apriori that the receiver
can handle it.  We started out with a request for a "canonical
master file format".  That is ascii and only ascii.  Yes lowest common
denominator.

> >Additionally you need to be aware of what types the nameservers
> >supports otherwise you should encode all non RFC 1034 records as
> >unknown types.  Unfortunately RFC 103[45] failed to specicify how
> >to encode unknown types in master files which makes it impossible
> >to have a master files that contains new types be readable by all
> >nameservers in existence.
>
> I’m well aware of this.  RFC 1034/1035 didn’t specify how you’d include
> write an RRTYPE they didn’t invent, not even the ones in "New DNS RR
> Definitions” RFC 1183 (October 1990, Experimental - ok, outside of RP
> non are in use and RP may or may not be).  It’s a shame RFC 1034/1035
> didn’t cover this (but there were many omissions, so it’s not surprising)
> but we deal with that.  The printed form of unknown record types is no
> different that the printed form of, say AAAA (in some use [wink], not in RFC
> 1034/1035).
>
> The point is to define a way to print data into a file.  And not even “a”
> way, but have a lowest common denominator and then more useful
> “adjectives.”  It doesn’t matter if parsers today “break” or 
> “fail” to
> consume a particular format - if that happens, operators can perform “tech
> refresh” if they need.
>
> PS - I want to emphasize that LDH is not a requirement for a domain name.
> (I think that point was made already.)  A label can have only LDH, it can
> have punydode, it can have arbitrary binary code.  Printing a label (as it
> is seen in a file) isn’t the same as converting a name for registration.

-- 
Mark Andrews, ISC
1 Seymour St., Dundas Valley, NSW 2117, Australia
PHONE: +61 2 9871 4742                 INTERNET: [email protected]

_______________________________________________
DNSOP mailing list
[email protected]
https://www.ietf.org/mailman/listinfo/dnsop

Reply via email to