Versioning of the standard vs. the software

[vaitkus]

I was wondering if the InChI developers make any distinction between the InChI specification version and the InChI software release version? Is the InChI specification always relayed only by the major version number (as is embedded in an InChI string) or does it follow the same versioning as the software releases, i.e. the latest InChI specification version is 1.07.04?

In other words, can the specification be assumed to be frozen as reported in the original InChI publications, or is it still evolving outside the original reported scope?

[nbehrnd]

To me, addressing this question is reasonable because of @fbaensch-beilstein's comment by August 15th to PR #107 reading

The 1B should mark InChIs generated with this new option as ‘still in beta status’ in the next release or in the first release that includes this feature. If successfully tested (by the community), the feature will (hopefully) be included in the standard InChI.

while commits leading to PR #107 so far edit .c and .h files only.

[fbaensch-beilstein]

This is a topic that has been discussed a lot recently.
As things stand, the software is being versioned numerically and the InChI string itself will remain with the prefix InChI=1S/ for everything that is included in the standard InChI and runs stably, so that there will be no differences between InChI strings generated with software v1.06 and software v1.07.x. The InChI=1B/ prefix is used for features that are still being tested, such as the new molecular inorganics functionalities (from #107).
If you have any further questions, please do not hesitate to ask.

[fbaensch-beilstein]

@vaitkus Have you found the answer you were looking for? If so, can we now close this matter?

[vaitkus]

Sorry for not responding sooner.

So if I understand correctly, InChI software releases may introduce new features (and maybe even InChI layers) to the non-standard InChI (e.g. the original publication/standard probably does not describe the 1B prefix), however, the standard InChI are guaranteed to remain the same (assuming the previous implementation did not have any serious issues)?

In other words, standard InChI generated by v1.07 will be parseable and correctly interpretable by any previous software release while the same is not true for non-standard InChI? But isn't the version number encoded in non-standard InChI (i.e. InChI=1/...) a bit ambiguous then?

[fbaensch-beilstein]

No worries, I just checked the open issues to see which ones could possibly be closed.

It's a little confusing, I admit. The prefix with 1 indicates that it is an InChI string of the first version. 1S stands for the standard InChI of the first version, which includes features that have been tested accordingly. 1B includes ‘beta features’ that are still in the open testing phase.
One premise is that InChIs within a version are not changed (unless they are incorrect and are corrected by a bug fix). Let's take the example of the new molecular inorganic features. Here, new functionalities are introduced that ‘correct’ InChIs for molecular inorganic structures. However, it is important that these changes do not alter organic structures or their InChIs.
It can therefore be assumed that, for example, the InChI for caffeine will not change throughout the entire first version.

Whether and when there will be an InChI with the prefix 2, i.e. a version 2, is still open.

[vaitkus]

Thank you, this clears up things a bit!