[24905] trunk/base/ChangeLog

[James Berry]

Hi Ryan,

On May 7, 2007, at 9:14 PM, Ryan Schmidt wrote:

> On May 7, 2007, at 22:33, source_changes at macosforge.org wrote:
>> +Release 1.4.40 (7-May-2007):
>> +
>> +    - Note the bump in version naming. To leave ourselves lots of  
>> room in our versioning
>> +      scheme, we've jumped from 1.4.3 to 1.4.40. The floating  
>> point represenation as
>> +      reported by port version (1.440) will still be the same;  
>> we're just interpreting
>> +      it differently.
>> +
>>      - variable tracing now works in a much better way and handles  
>> unsets properly.
>>        Similarly, ${option}-delete now works better. Depends  
>> validation no longer
>>        attempts to validate when the variable is unset.  
>> Additionally, the validation
>
> Out of curiosity, why are we using these weird version numbers?
>
> The standard in the unix software world (which is the world of  
> software that MacPorts installs so it's not unreasonable to make  
> the comparison) would be that you have 1.4.0, 1.4.1, 1.4.2 etc up  
> to 1.4.9, then 1.4.10, 1.4.11, etc. Simply count up the last  
> number, until it's time to increment the middle number and reset  
> the last number back to zero.
>
> But this change log entry seems to describe a rather different  
> version strategy for MacPorts. In particular, it seems to mean that  
> the version after 1.4.9 must be 1.5.0; that there is no room for a  
> 1.4.10 or any larger version (unless you say that "1.491"  
> corresponds to 1.4.10, "1.492" to 1.4.11, etc., which would be a  
> mess). Is there a reason why we are intentionally limiting  
> ourselves with this numbering scheme? Is this a reason for wanting  
> to do fewer releases -- so that we don't have to prematurely come  
> to version 1.5.0 which might then not be substantially different  
> from 1.4.9?

Actually, the change mentioned above was to give ourselves lots more  
version numbers so that we _don't_ run out of numbers. The problem is  
that our versioning is governed by a single floating point number,  
which needs to increase from one release to the next, and which maps  
to our human readable number. So 1.4.3 is 1.430000... If we had  
continued on our path, we would soon have got to 1.4.9 = 1.49 and  
we'd be at the end of our 1.4 release.

So our new scheme essentially allocates another digit, so now we can  
go 1.4.40, 1.4.41, 1.4.42, ... 1.4.51 ... 1.4.68, ... 1.4.79 etc.  
It's not ideal, but the floating point form isn't ideal either; at  
some point we'll want to change that.

James