[openssl-project] A proposal for an updated OpenSSL version scheme (v2)

Tim Hudson tjh at cryptsoft.com
Fri Sep 21 13:28:01 UTC 2018


Now I get the conceptual issue that Richard and Matt are differing on - and
it is about actually replacing OpenSSL's versioning concept with semantic
versioning compared to adopting semantic versioning principles without
actually being precisely a semantic version approach.

The whole concept of semantic versioning is that you define *precisely *what
you mean by a version.
Everywhere you have the concept of a version you must use the semantic form
of a *version encoding*.

That is the X.Y.Z[-prerelease][+buildmeta] format that is documented along
with the rules for X.Y.Z in terms of your public API.
And all other information about a version goes into prerelease and into
buildmeta.
Both prerelease and buildmeta are allowed to be a sequence of dot separated
alphanumerichyphen combinations.

This is the point of semantic versioning. All versions for all products are
all represented with the same sort of concepts and you know what the rules
are for the numeric X.Y.Z handling and the parsing rules for prerelease and
buildmeta.

Our concepts of versioning within OpenSSL if expressed in semantic form
MUST fit into this approach.
No prefixes. No suffixes. Not special additional encoding The idea is
consistency.

When dealing with API issues you only ever need to see X.Y.Z for any code
related testing - it precisely identifies a point in time API release.
There should never be any code ever that varies that requires looking at
prerelease or buildmeta in order to perform any action in terms of the code.

That maps to our concept of OPENSSL_VERSION_NUMBER

For the human reporting we should have the fill concept which is a text
string - and that should be OPENSSL_VERSION_TEXT and that means not having
anything else within the text other than the actual semantic version.
The syntax of the semantic version is fixed.

If you want to keep the concept of a build date in the macro you are
calling the version then it must be encoded in that format - or you move
that concept out of the macro that is the version.

Tim.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://mta.openssl.org/pipermail/openssl-project/attachments/20180921/d32fbd27/attachment.html>


More information about the openssl-project mailing list