GNU bug report logs - #57237
b2sum does not support '-a' options found at https://www.blake2.net/

Previous Next

To reply to this bug, email your comments to 57237 AT debbugs.gnu.org.

Toggle the display of automated, internal messages from the tracker.

Message #5 received at submit <at> debbugs.gnu.org (full text, mbox):

From: "Robert E. Novak" <sailnfool <at> gmail.com>
To: bug-coreutils <at> gnu.org
Subject: b2sum does not support '-a' options found at https://www.blake2.net/
Date: Mon, 15 Aug 2022 20:15:51 -0700

as a result, the Gnu version does not support blake2s (for smaller 
digests) and blakes2bp for higher performance on multicore systems.

Yes, I know about Blake3, but there are many reasons to support older 
hash algorithms.

The real problem is that you have co-opted the b2sum binary so that the 
testing required to find out if is system's b2sum application is the 
open source b2sum from https://www.blake2.net/ or the Gnu coreutils.  I 
am hopeful that a similar problem is not introduced with a coreutils 
version of b3sum?  Will you implement the C language single thread 
version (lower performance) or the parallel merkle tree rust implementation?

Since you introduce incompatible differences in the implementation, the 
least that you could do is to rename the Gnu Coreutils b2sum to gnub2sum 
so that applications that require different command line semantics do 
NOT Have to go through machinations to find all installed versions of 
b2sum on a system in order to select the correct invocation semantics.  
Just my $0.02 worth, but I am trying to rationalize the world of 
cryptographic hash algorithms.  I have two blogs that reference this on 
linkedin ( * & ** ) so that you can understand part of the reason why 
this will become more important over time.

I realize that Gnu has a long tradition of implementing the Gnu version 
of commands and that in many cases the Gnu versions have become the "de 
facto" standards.  However this does not happen if you don't support the 
semantics of the commands that you are replacing.

I have been using Unix/Linux since 1974 (Arpanet node #6 at Urbana, 
Illinois) and I would never have made the transition to Linux if there 
were less semantic consistency over time.

If indeed you had implemented a superset of the Blake2 semantics, there 
would be no cause for concern.

* 
https://www.linkedin.com/pulse/canonical-cryptographic-hash-encoding-robert-e-novak/?trackingId=gjy%2FJwsjnJaviUN2ZYtuqw%3D%3D

This is a preliminary version and I hope to release a second version 
after further development.

** 
https://www.linkedin.com/pulse/thoughts-pragmatic-one-time-pads-encryption-robert-e-novak/?trackingId=3AUdLAGWSsDMYYuCINZuVA%3D%3D

Message #8 received at 57237 <at> debbugs.gnu.org (full text, mbox):

From: Pádraig Brady <P <at> draigBrady.com>
To: "Robert E. Novak" <sailnfool <at> gmail.com>, 57237 <at> debbugs.gnu.org
Subject: Re: bug#57237: b2sum does not support '-a' options found at
 https://www.blake2.net/
Date: Tue, 16 Aug 2022 13:44:08 +0100

On 16/08/2022 04:15, Robert E. Novak wrote:
> as a result, the Gnu version does not support blake2s (for smaller
> digests) and blakes2bp for higher performance on multicore systems.
> 
> Yes, I know about Blake3, but there are many reasons to support older
> hash algorithms.
> 
> The real problem is that you have co-opted the b2sum binary so that the
> testing required to find out if is system's b2sum application is the
> open source b2sum from https://www.blake2.net/ or the Gnu coreutils.  I
> am hopeful that a similar problem is not introduced with a coreutils
> version of b3sum?  Will you implement the C language single thread
> version (lower performance) or the parallel merkle tree rust implementation?
> 
> Since you introduce incompatible differences in the implementation, the
> least that you could do is to rename the Gnu Coreutils b2sum to gnub2sum
> so that applications that require different command line semantics do
> NOT Have to go through machinations to find all installed versions of
> b2sum on a system in order to select the correct invocation semantics.
> Just my $0.02 worth, but I am trying to rationalize the world of
> cryptographic hash algorithms.  I have two blogs that reference this on
> linkedin ( * & ** ) so that you can understand part of the reason why
> this will become more important over time.
> 
> I realize that Gnu has a long tradition of implementing the Gnu version
> of commands and that in many cases the Gnu versions have become the "de
> facto" standards.  However this does not happen if you don't support the
> semantics of the commands that you are replacing.
> 
> I have been using Unix/Linux since 1974 (Arpanet node #6 at Urbana,
> Illinois) and I would never have made the transition to Linux if there
> were less semantic consistency over time.
> 
> If indeed you had implemented a superset of the Blake2 semantics, there
> would be no cause for concern.
> 
> *
> https://www.linkedin.com/pulse/canonical-cryptographic-hash-encoding-robert-e-novak/?trackingId=gjy%2FJwsjnJaviUN2ZYtuqw%3D%3D
> 
> This is a preliminary version and I hope to release a second version
> after further development.
> 
> **
> https://www.linkedin.com/pulse/thoughts-pragmatic-one-time-pads-encryption-robert-e-novak/?trackingId=3AUdLAGWSsDMYYuCINZuVA%3D%3D


Discussion on the initial GNU coreutils implementation,
including dropping of the -a interface was discussed at:

  https://lists.gnu.org/archive/html/coreutils/2016-10/msg00007.html
  https://lists.gnu.org/archive/html/coreutils/2016-11/msg00000.html

We broached keeping coreutils simpler using just blake2b with blake2 folks,
and there was general agreement.  So essentially there were design negotiations
in the threads above to present the interface appropriate to most users
in the GNU coreutils b2sum implementation,
which would become the standard variant available.

Which systems are installing the reference version?
I presume this is an infrequent issue?
I would think the onus on the above systems would be to install the reference
version under a different name.

Note the GNU coreutils digest tools were refactored recently
around a single `cksum -a ...` tool, and there will be no
new separate digest specific tools in future (like b3sum etc.).

thanks,
Pádraig

Previous Next

GNU bug tracking system
Copyright (C) 1999 Darren O. Benham, 1997,2003 nCipher Corporation Ltd, 1994-97 Ian Jackson.