Re: Empty attribute values

From: Paul Rohr (paul@abisource.com)
Date: Thu May 30 2002 - 12:20:46 EDT

  • Next message: David Chart: "Re: Still can't build HEAD"

    At 02:05 PM 5/27/02 -0400, Patrick Lam wrote:
    >I stumbled across the following comment:
    >
    > /*
    > Jeff, I weakened the following assert because we
    > *want* to represent no tabstops as an empty string.
    > If this isn't safe, let me know. -- PCR
    > */
    > UT_ASSERT(p[1] /* && *p[1] */); // require value for each name
    >
    >It seems that we have a bunch of other empty props now too, like
    >TF's new 'display' property.

    Yeah. That's my comment. It's very un-CSS-like to not have a value for a
    given name, and if I'd thought of a better way to spec the tabstops value, I
    wouldn't have made the change.

    Likewise I empathize with TF's decision. For good reasons, we don't use
    CSS's box formatting model, so there may not *be* an appropriate setting for
    our default behavior:

      - both inline and block from CSS1 are clearly wrong
      - both compact and run-in from CSS2 are also wrong

    Thus, in both cases, there's a weakly sufficient reason to *not* have a
    value for those specific property names.

    >Now, I've been working on some field code (bug: start typing in
    >italics and insert a field; the field does not appear in italics!)
    >and in fixing this bug, I get the corresponding assert in the
    >attributes code.
    >
    >That's because a field has an empty string "" in its param attribute.
    >Do we really want that? If so, I'll weaken the attribute assert too.

    Ick. That sounds ugly to me too. I doubt that this is by design, but you
    should check with Martin to confirm.

    Paul



    This archive was generated by hypermail 2.1.4 : Thu May 30 2002 - 12:24:20 EDT