Paul Eggert via tz wrote in
<[email protected]>:
|On 2026-03-06 18:08, Robert Elz via tz wrote:
...
|> where indeed PT would not be usable, but we have no reason to care about
|> that
|
|Unfortunately we do. See Internet RFC 9636 section 3.3
One more of those RFCs that should never have been written!
And if standardizing a binary file format of some library, then
some standardized binary format should have been used. I was
asking for CBOR by then, granted for the over-the-net tz stuff
that does not fly, and, of course, only as a suggestion, yaml or
even (i hate it for several mostly good Unicode reasons) json or
anything that is standardized and has schema support.
Ie, all the same data, all the same format.
The "Interoperability Considerations" is self-speaking against the
binary "upward-compatible" of the introduction, and mostly
everywhere the project code is used, accompanying the data, as
i see it, otherwise the considerations have to be considered.
Right.
Btw personally i am hoping that the Go TZif reader that you link
from your own homepage, that its comments form the critique i
would state, especially that ~"and now we can form the actual
data" is indeed Monty Pythonesque. And then you even need bug
workarounds for certain implementations.
I would vote for adding another variant, in a standardized
container format, eventually growing out the old one, installed
like /etc/localtime.X where X is accordingly.
The RFC speaks about "obsolescent version-1-only readers MUST", so
the RFC implies code changes are doable on even those obsolescent
version-1-only readers. (Shudder.)
...
|Although we could relax these restrictions in later RFCs, that's not
...
Or drop those altogether. They have no value, and never had,
really, in my opinion, except for possibly specifying a data
schemata. Soon all readers are obsolescent, and the new RFC would
have to add more paragraphs on "placeholder version XY
datablocks".
There is still over a decade until 64-bit time is really needed.
Turn to localtime.X where X is, i am guessing, JSON.
(Despair here.)
(Btw, and now off-topic, i know a former member of FSF Europe
who is prowd of a personal signature of Mr. Stallman, and who is
in the process of developing
KEKS is compact, deterministic, concise and streaming binary
serialisation format. It is aimed to be lightweight in terms of CPU,
memory, storage and codec implementation size usage. It supports wide
range of data types, making it able to transparently replace JSON.
KEKS/Schema is a schema definition format for describing data
structures validation steps.
etc with already implementations in C, Go, python3, tcl, if i get
this right. git://git.cypherpunks.su/keks.git.
Since you seem to be interested in good small things, and
surrounding the FSF universe, like Diaz's lzip.)
...
|something we could reasonably expect all downstream TZif readers and
|producers to implement by November, so for now the restrictions do
|impinge on what we can put into America/Vancouver for this issue.
|
|Besides, alphabetic time zone abbreviations are problematic given their
|ambiguity and lack of standardization, and so I doubt whether it'd be
|wise to relax the restrictions and encourage use of shorter
|abbreviations like "PT" or (shudder!) "O". We have trouble enough as it
|is with these things.
|
|I quite agree, and POSIX.1-2024 started the ball rolling on getting rid
|of that garbage. Alas the job is still incomplete. (This is veering off
|into a different subject of course.)
Drop it altogether. There are incompatibilities, there are bugs,
the number of actively developed TZif readers is not that large,
as far as i know, and those which do often use languages which
ship infrastructure to go JSON/X out-of-the-box, and easily.
And with automatic schema you then also get *data content
verification* for free, as opposed to only binary data structure.
(As in, "how many readers perform really proper verification?")
And with a filename with regular file extension you integrate in
"the normal" mime.types approach much better, ie, RFC 9636's
"When a TZif file is used in a MIME message entity, it SHOULD be
indicated by one of the following media types" requires "magic"
logic (aka file content inspection, as opposed to only file
extension checks via eg /etc/mime.types) that many applications do
not support, not even those that make use of the monstrous (imho)
freedesktop.org/xdg/shared-mime-info it seems
$ pkginfo -l shared-mime-info | grep -i tz
usr/share/mime/application/x-tzo.xml
and that one unfortunately cannot get rid of
$ prt-get dependent shared-mime-info
gobject-introspection
$ prt-get dependent gobject-introspection
glib-introspection
$ prt-get dependent glib-introspection
gdk-pixbuf
at-spi2-core
harfbuzz
Sigh. Granted file(1) of NetBSD's Zoulas can
$ file /usr/share/zoneinfo/Iran
/usr/share/zoneinfo/Iran: timezone data (slim), version 2, no gmt time flags,
no std time flags, no leap seconds, 71 transition times, 6 local time types, 28
abbreviation chars
pretty impressive, but wait
$ file --mime-type /usr/share/zoneinfo/Iran
/usr/share/zoneinfo/Iran: application/octet-stream
or what
$ file --mime-encoding /usr/share/zoneinfo/Iran
/usr/share/zoneinfo/Iran: binary
Noone ever cared. (But even if, i for example have my own
mime.types ie based on extensions, and it will prowdly support
things like
# built-in
mimetype application/x-lzma lzma
# built-in (not for sending mail)
mimetype ?only-handler application/x-tar-lzma tar.lzma
with the next release.)
"Get rid of that garbage", yes. (RFC, not readers -- i have not
really looked at their code. But code there is, sometimes quite
a bit.)
And tzdata.si is great.
Thank you.
--End of <[email protected]>
--steffen
|
|Der Kragenbaer, The moon bear,
|der holt sich munter he cheerfully and one by one
|einen nach dem anderen runter wa.ks himself off
|(By Robert Gernhardt)