Do you think the proposed 2.3.1 that depends on uimaj 2.9.0
could instead depend on uimaj 2.9.0 or 2.10.0?

-M


On 5/17/2017 3:15 PM, Richard Eckart de Castilho wrote:
> On 17.05.2017, at 20:41, Marshall Schor <[email protected]> wrote:
>> 2 thoughts:
>>
>> 1) if your component doesn't require 2.10.0, but works with a range of 
>> versions,
>> including 2.10.0, and
>>
>>  - your component doesn't "include" core uima 2.10.0, then I think the 
>> version
>> numbering can be independent.
> We could turn uimaFIT into a module of uimaj ;) Then we'd have
> only once release cycle. When uimaFIT was contributed, I was expecting that
> uimaFIT would be released more often than uimaj-core and I wanted to be able
> to have these faster release cycles. However, in reality, it turns out this
> was wishful thinking. At least at the moment, I rather manage to do less
> releases than Marshall.  
>
>> 2) I suggest we be customer-focused, and try to have our version numbering
>> communicate useful information.  So, to me, if we have a bug-fix release that
>> doesn't change customer-facing APIs except for perhaps adding new signatures
>> (which are very unlikely to collide with user methods) , I would treat this 
>> as a
>> 3-rd digit increment, because that communicates best what an upgrader ought 
>> to
>> expect.
> I didn't feel comfortable upgrading the uimaj-core dependency to 2.10.0 and 
> then
> only to increase the last digit. If felt that it would have been less 
> customer-friendly
> to do that then to increase the middle number.
>
> Anyway, it was probably a bad idea anyway as Peter has pointed out.
>
> I'll do prepare a 2.3.1 that depends on uimaj-core 2.9.0.
>
> -- Richard

Reply via email to