Bruno Haible via Gnulib discussion list <[email protected]> writes:

> Simon Josefsson wrote:
>> Having to specify the hash algorithm is tedious for users...  maybe we
>> should use the following style instead, and merely suggest '--check':
>
> I disagree:
> 1) What is tedious for users is to understand that they are supposed to
>    copy-and-paste a line from the documentation into an 'echo' statement.
>    Compared to that, pasting one or two more options is trivial.
> 2) You are suggesting to change the documentation to match a tool. It
>    would be better to change the tool to match the style used in the
>    documentation.

Yes, but the current line does not have any information about the hash
algorithm:

+6kYv60oHv7kXpkJM2vUlADWNmh62nus1xA80bJJiJEs=  gnulib-20260109.bundle

vs

SHA3-256 (gnulib-20260109.bundle) = 6kYv60oHv7kXpkJM2vUlADWNmh62nus1xA80bJJiJEs=

so there is no reliable way to change the tool to guess the first format
is for SHA3-256.  Running all hash algorithms in parallel and accept one
that matches would be insecure.

I've never been a fan of the second format above, but I think this
argument is compelling to me.  Formats to convey hash values should
include the hash algorithm identifier, otherwise people could be tricked
into using the wrong algorithm.  I don't see any immediate problem with
our release announcements, because the hash algorithm is clear from
context in them, but this is also about establishing a good pattern.

To avoid '-a FOO' the input format need to mention which hash algorithm
to use, and we have that in the default behaviour of cksum with the
tagged format.  I don't think having the user supply the algorithm will
bring any additional security information, but just make things harder.

/Simon

Attachment: signature.asc
Description: PGP signature

Reply via email to