[plt-scheme] proposal for indicating planet package version numbers

From: Neil Van Dyke (neil at neilvandyke.org)
Date: Sat Aug 30 22:16:59 EDT 2008

Sorry, I misread, trying to read sideways. :)

The concern I have with marginalizing the real version number comes from 
the fact that I try to develop portable Scheme libraries that are used 
on many different Scheme implementations.  I do this for practical 
reasons, so that people can invest in Scheme and move between 
implementations for different requirements.  (Say, PLT by default, 
something else if one needs to run in a JVM, use massive numbers of 
threads, run on a microcontroller, start using a new research Scheme 
implementation, run in a different Scheme that is already embedded in an 
application, or...)

I believe that the other implementation-specific packagings of my 
libraries all use the canonical version number.  So when I say "Foo 
1.3", everyone is talking about the same thing.  PLT is my favorite 
all-around Scheme, but I would be sad if I had to say "Foo 1.3 (which is 
2.1 to PLT users)" and "oh, all this time you meant *PLT* 2.1, which 
means you're actually using... hm... Foo 1.3, and you really need Foo 
2.1, which is... hm... PLT 3.0..."

I'm not saying that my needs alone should affect PLaneT design 
decisions, but I think that some of my needs might be shared by other 
PLT users, now or in the near future.

Just to step back a moment, I'm not sure that this needs to be a big 
issue.  Using "." in the PLT package version number on the Web page made 
no sense to me from the start, and this use of "." is the only problem.

Robby Findler wrote at 08/30/2008 09:50 PM:
> Oh wait -- I think maybe I wasn't clear. I'm proposing keeping only
> the planet version numbers (and getting rid of the other number). The
> planet version number is the only number planet is concerned with --
> the other number is just decoration, as far as planet is concerned. I
> don't see why it needs to be as prominent as it is. If it is that
> important to your package, it is easy to put it in the one-line
> description that shows up all over the place.
> Robby
> On Sat, Aug 30, 2008 at 8:47 PM, Neil Van Dyke <neil at neilvandyke.org> wrote:
>> If that would work for PLaneT, that would be excellent.
>> It seems like there's definitely value in having PLaneT know for each
>> version with which previous versions the version is API backward-compatible.
>>  The PLaneT version numbers seem like an easy, if imperfect, way to do that
>> in most cases.  I don't yet know how well that will work in practice with
>> the large number of small, interconnected libraries that I'd like to do (in
>> which a developer would "require" one library and typically pull in several
>> dependencies as well).
>> Robby Findler wrote at 08/30/2008 09:36 PM:
>>> In my opinion, the non-PLaneT version number should just go away
>>> (except possibly being mentioned in the documentation or release notes
>>> for packages where the planet package is a re-packaging of something
>>> else). For most packages, the number is irrelevant and, as you point
>>> out, for others it is confusing.
>>> Would that also satisfy you?
>>> Robby
>>> On Sat, Aug 30, 2008 at 8:27 PM, Neil Van Dyke <neil at neilvandyke.org>
>>> wrote:
>>>> To reduce confusion between PLaneT "package versions" and the version
>>>> numbers of software packages that happen to be packaged as PLaneT
>>>> packages... I'd like to strongly suggest that instead of having the
>>>> PLaneT
>>>> Web pages say this:
>>>> Package
>>>> version   Version
>>>> 1.1       0.3
>>>> that they instead say this:
>>>> Package
>>>> version   Version
>>>> (1 1)     0.3
>>>> Basically, I want to never, ever see "." in a PLaneT package version
>>>> number.  "." is already used in the version numbers of most software
>>>> packages ever written, which means it terrific for distinguishing the two
>>>> kinds of version numbers.  TThe sexp format proposed above is also a good
>>>> mnemonic for the "Scheme" flavor of version number, as opposed to the
>>>> "normal" flavor.  The sexp format also fits the different ways that
>>>> PLaneT
>>>> package version numbers are used, better than a string with dots in it
>>>> would, and "require" forms don't use dots anyway.
>>>> If someone could make this change, I also suggest clearer terminological
>>>> conventions, such as saying either "PLaneT Version" and "Software
>>>> Version",
>>>> and never simply saying "version".  Or invent a new word for "PLaneT
>>>> version" -- something like "verid", which can be selected in "require"
>>>> forms
>>>> and such by a "verspec" form, only perhaps not so dorky-sounding.
>>>> I think I suggested this before, but I don't recall seeing a rationale
>>>> for
>>>> doing it the current way.
>>>> --
>>>> http://www.neilvandyke.org/
>>>> _________________________________________________
>>>>  For list-related administrative tasks:
>>>>  http://list.cs.brown.edu/mailman/listinfo/plt-scheme

Posted on the users mailing list.