[r6rs-discuss] [Formal] Version numbers are broken in 5.91

[Michael Sperber]

---
This message is a formal comment which was submitted to formal-comment at r6rs.org, following the requirements described at: http://www.r6rs.org/process.html
---

Submitter's email address: sperber at deinprogramm.de
Type of issue: Defect 
Priority: Major
R6RS component: Libraries
Version of the report: 5.91
One-sentence summary of the issue: The treatment of version numbers
needs to be changed

This post by Marcin Kowalczyk describes the issue:

http://lists.r6rs.org/pipermail/r6rs-discuss/2006-September/000013.html

Quoting the post:

> 6.1.
> 
> As far as I understand it, it's possible to specify conditions in
> version references only in each subversion field separately.
> 
> For example it's not possible to say "version (3 n) for n>=2 or
> version (m n) for m>=4" (i.e. "3.2 or newer"), nor even "(3 2 . m)
> for any tail m".
> 
> Perhaps I misunderstood how version numbers would typically be used in
> practice. But wouldn't it be better if "and" and "or" were applicable
> to version conditions as a whole, rather than individual subversion
> fields?

How to fix the issue:

While the above disclaimer applies, I wrote the original versioning
spec, and this seems like a clear omission.

Here's how I'd like to fix it:

<version reference> -> <empty>
                     | <version spec>     ; new

<version spec> -> (<subversion reference> ...)
                | (and <version spec> ...)
                | (or <version spec> ...)
                | (not <version spec> ...)

This would allow saying:

(or (3 (>= 2))
    ((>= 4)))

which would match:

3.2, 3.3, 3.2.1, 3.5.2, etc., 4, 4.0, 4.1.1, etc.

I believe the dual occurrence of `and', `or', and `not' is
unambiguous, as crossing from <version spec> to <subversion reference>
always involves an additional open parenthesis, and there's no way
back up.