# Re: [gcpm] border-parts

```On Nov 5, 2008, at 8:53 AM, Håkon Wium Lie wrote:

>
>> Well then, I would like to see it (or something like it) be put back
>> in, to make this proposal more widely usable, and to help satisfy a
>> larger demand (without having a separate proposal for "dash-parts"
>> that is almost identical). What about about a keyword that makes the
>> entire pattern into a repeating pattern, instead of just sections of
>> it? like { border-parts: 20px 10px 0 1fr, repeat }
>
> This is idendtical to 'border-parts: repeat(20px 10px 0 1fr)' in the
> previous proposal, it seems.

Correct. Except that, if I understand correctly, one of the things
that made it complicated was when you stung it together with other
parts, like this:

border-parts: 1fr 5px repeat(20px 10px 0 1fr);

So my change to help simplify is to say that "repeat" would be for the
whole string of lengths, and not for substrings of lengths.

> The problem is the competition that appears between repeat and
> fractions. We think of fractions as eating up the leftover space.
> However, 'repeat' is also very good at eating leftovers and this makes
> fractions go towards 0.

True, but with repeated patterns I would really like to be able to
have dash patterns that line up on the corners nicely, preferably in a
way that was symmetrical on both ends. And that is where I would see
the flex space being used in this case, to add a little space that
ensures my corners are square (or round, if it is a dotted pattern
instead of dashed), and that they are not occupied by a space, and the
the right corners are mirrors of the left corners. Maybe there would
be another way to do that, but my thought was that the flex unit (or
fraction unit) would be used that way. If no fractional units are used
then the dashes just go clockwise around each side. But if they are
used, then the length of the pattern is stretched there, so that the
length of the pattern is evenly divisible into the length of the
border edge.

> When used separately, both mechanisms have merit.
>
> But I don't see them working well together on the same property.