Re: uuid and RFC4122

Re: uuid and RFC4122

[Chris Hofstaedtler]

Hello Michael,

* Michael Richardson <mcr@sandelman.ca>:
> Hi, thanks for your work on debian packages.
> I'm contacting you wearing my IETF uuidrev WG co-chair on.
> 
> The https://datatracker.ietf.org/wg/uuidrev/documents/ WG was chartered to
> update RFC4122 last summer, and we expect to start a WGLC on RFC4122bis in
> the next few weeks.  This is just a heads up for now.
> 
> Some comments, even they are just "Looks Good to Me" from the util-linux
> folks would be awesome.

Thanks for reaching out, but from the Debian side I think we have
little to comment on (mostly for lack of expertise).

I've CC:ed the upstream mailing list, in the hope that upstream
util-linux folks have comments.

Best,
Chris

Re: uuid and RFC4122

[Theodore Ts'o]

On Sun, Feb 19, 2023 at 06:01:16PM +0100, Chris Hofstaedtler wrote:
> Hello Michael,
> 
> * Michael Richardson <mcr@sandelman.ca>:
> > Hi, thanks for your work on debian packages.
> > I'm contacting you wearing my IETF uuidrev WG co-chair on.
> > 
> > The https://datatracker.ietf.org/wg/uuidrev/documents/ WG was chartered to
> > update RFC4122 last summer, and we expect to start a WGLC on RFC4122bis in
> > the next few weeks.  This is just a heads up for now.
> > 
> > Some comments, even they are just "Looks Good to Me" from the util-linux
> > folks would be awesome.
> 
> Thanks for reaching out, but from the Debian side I think we have
> little to comment on (mostly for lack of expertise).
> 
> I've CC:ed the upstream mailing list, in the hope that upstream
> util-linux folks have comments.

Hi,

I originally wrote the uuid implementation in util-linux.  Originally,
it was first in e2fsprogs, and then later the library was transferred
to util-linux at least for most Linux distributions.  Technically
speaking, however, libuuid is still shipped in the e2fsprogs sourecs
and that's the version which is used by Android and some non-Linux
OS's including NetBSD and Darwin (MacOS may have forked the library,
though, so if I shipped updates in e2fsprogs how quickly it would get
picked up by Apple.)  Libuuid is mostly in maintenance mode, actually;
the sort of changes that have landed recently have been to support the
getrandom(2) system call.

I'm not sure what the context is for this e-mail thread, or what
you're asking for in terms of an LGTM?  Were you looking for comments
on draft-ietf-uuidrev-rfc4122bis?  Or were you hoping to see if folks
might be interested in implementing the new UUID versions in
draft-ietf-uuidrev-rfc4122bis so the libuuid in Linux might support
those new variants?

Cheers,

					- Ted

P.S.  Hi Michael; long time no chat --- I haven't been to an IETF
meeting in many years!)

Re: uuid and RFC4122

[Karel Zak]

On Sun, Feb 19, 2023 at 08:39:03PM -0500, Theodore Ts'o wrote:
> On Sun, Feb 19, 2023 at 06:01:16PM +0100, Chris Hofstaedtler wrote:
> > Hello Michael,
> > 
> > * Michael Richardson <mcr@sandelman.ca>:
> > > Hi, thanks for your work on debian packages.
> > > I'm contacting you wearing my IETF uuidrev WG co-chair on.
> > > 
> > > The https://datatracker.ietf.org/wg/uuidrev/documents/ WG was chartered to
> > > update RFC4122 last summer, and we expect to start a WGLC on RFC4122bis in
> > > the next few weeks.  This is just a heads up for now.
> > > 
> > > Some comments, even they are just "Looks Good to Me" from the util-linux
> > > folks would be awesome.
> > 
> > Thanks for reaching out, but from the Debian side I think we have
> > little to comment on (mostly for lack of expertise).
> > 
> > I've CC:ed the upstream mailing list, in the hope that upstream
> > util-linux folks have comments.
> 
> Hi,
> 
> I originally wrote the uuid implementation in util-linux.  Originally,
> it was first in e2fsprogs, and then later the library was transferred
> to util-linux at least for most Linux distributions.  Technically
> speaking, however, libuuid is still shipped in the e2fsprogs sourecs
> and that's the version which is used by Android and some non-Linux
> OS's including NetBSD and Darwin (MacOS may have forked the library,
> though, so if I shipped updates in e2fsprogs how quickly it would get
> picked up by Apple.)  Libuuid is mostly in maintenance mode, actually;
> the sort of changes that have landed recently have been to support the
> getrandom(2) system call.
> 
> I'm not sure what the context is for this e-mail thread, or what
> you're asking for in terms of an LGTM?  Were you looking for comments
> on draft-ietf-uuidrev-rfc4122bis?  Or were you hoping to see if folks
> might be interested in implementing the new UUID versions in
> draft-ietf-uuidrev-rfc4122bis so the libuuid in Linux might support
> those new variants?

Note that libuuid in util-linux is actively maintained and improved.

It already supports MD5 and SHA1 based UUIDs, and I think we can add
the new versions v6 (Reordered Gregorian time-based), v7 (Unix Epoch
time-based) and v8 (custom SHA256/SHA512 based) in the near feature.

CC also to Michael, who worked on libuuid in last years; not sure
but maybe v6, which improves use in databases, could be attractive for
SAP, see:
https://datatracker.ietf.org/doc/draft-ietf-uuidrev-rfc4122bis/

    Karel

-- 
 Karel Zak  <kzak@redhat.com>
 http://karelzak.blogspot.com

Re: uuid and RFC4122

[Michael Richardson]

[-- Attachment #1: Type: text/plain, Size: 1909 bytes --]


Theodore Ts'o <tytso@mit.edu> wrote:
    > I originally wrote the uuid implementation in util-linux.  Originally,
    > it was first in e2fsprogs, and then later the library was transferred
    > to util-linux at least for most Linux distributions.  Technically
    > speaking, however, libuuid is still shipped in the e2fsprogs sourecs
    > and that's the version which is used by Android and some non-Linux
    > OS's including NetBSD and Darwin (MacOS may have forked the library,
    > though, so if I shipped updates in e2fsprogs how quickly it would get
    > picked up by Apple.)  Libuuid is mostly in maintenance mode, actually;
    > the sort of changes that have landed recently have been to support the
    > getrandom(2) system call.

Good to hear, thanks for the reply.

    > I'm not sure what the context is for this e-mail thread, or what
    > you're asking for in terms of an LGTM?

Read the document, or just the diff:

  https://author-tools.ietf.org/diff?doc_1=rfc4122&doc_2=draft-ietf-uuidrev-rfc4122bis-02

    > Were you looking for comments
    > on draft-ietf-uuidrev-rfc4122bis?

Email uuidrev@ietf.org ... we are especially interested in whether we have
left out/obsoleted some use case that you presently have, whether the new
version types are useful to you.
Do you use or care about MD5 or SHA1 based UUIDs?
We have proposed using namespace based UUIDs for future hash methods.

    > Or were you hoping to see if folks
    > might be interested in implementing the new UUID versions in
    > draft-ietf-uuidrev-rfc4122bis so the libuuid in Linux might support
    > those new variants?

No, I think that if you had a new use case/need, you'd have implemented
something already :-)


--
Michael Richardson <mcr+IETF@sandelman.ca>   . o O ( IPv6 IøT consulting )
           Sandelman Software Works Inc, Ottawa and Worldwide





[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 515 bytes --]

Re: uuid and RFC4122

[Michael Richardson]

[-- Attachment #1: Type: text/plain, Size: 1060 bytes --]


Hello again!

Michael Richardson <mcr@sandelman.ca> wrote:
    > Hi, thanks for your work on debian packages.  I'm contacting you
    > wearing my IETF uuidrev WG co-chair on.

    > The https://datatracker.ietf.org/wg/uuidrev/documents/ WG was chartered
    > to update RFC4122 last summer, and we expect to start a WGLC on
    > RFC4122bis in the next few weeks.  This is just a heads up for now.

We have started the WGLC today.
It goes until June 9th.  The document is still open for comments,
particularly of an editorial nature.
Did something confuse you? our fault, let us know.

We included a way to do SHA-2 based uuids into the v8 namespace based uuids,
which we think is future proof, and won't burn version numbers the way that
MD5 and SHA1 derived uuids did.

You may find this diff to be useful:

https://author-tools.ietf.org/iddiff?url1=draft-ietf-uuidrev-rfc4122bis-01&url2=draft-ietf-uuidrev-rfc4122bis-03&difftype=--html

Your comments to uuidrev@ietf.org, or to https://github.com/ietf-wg-uuidrev/rfc4122bis/issues
(or even send a PR)





[-- Attachment #2: signature.asc --]
[-- Type: application/pgp-signature, Size: 511 bytes --]