On 27/05/2014 17:12, Bruno Le Floch wrote:
> On 5/27/14, Joseph Wright <[log in to unmask]> wrote:
>> On 27/05/2014 16:58, Bruno Le Floch wrote:
>>> I tend to agree with you that we should not let \__dim_strip_pt:n and
>>> \__dim_strip_bp:n accept arbitrary junk.  That's easy to change, as
>>> they are only used internally.  On the other hand, the public
>>> \dim_to_pt:n and \dim_to_unit:n are documented as allowing multiple
>>> units, for the use-case Joseph describes.
>>
>> If you look over the current actual uses in the LaTeX kernel, they are
>> focussed purely on some aspects of font use. I suspect we'd be better,
>> long-term, to allow this sort of thing if at all only at the interface
>> layer, and to have all of the 'general' code stuff expect 'well defined'
>> dimensions. Means a doc change, but I feel an reasonable one.
> 
> My worry is not about doc changes, but about people who might be using
> \dim_to_pt:n "in the wild".  We'd be breaking code.  Can we have a
> deprecation period?
> 
> Bruno

A reasonable point, but not always possible when we make an 'in place'
change (which are rare). Here, a sweep over TL2014 shows no uses outside
of the kernel code: I'd hope the impact will be very low.
-- 
Joseph Wright