<!DOCTYPE html>
<html>
  <head>
    <meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
  </head>
  <body>
    <p>Maybe this is the most useful thing I can say at this point:</p>
    <p>Almost all aspects of the various proposed extensions are tied <i>very
        closely</i> to existing parts of the specification. Something
      was 16 bit, now it's 24 bit. Cubic Beziers are added to the glyf
      table, but using an encoding very close to quadratic Beziers. New
      tables are added on the model of existing tables but can be
      longer. And so forth.<br>
    </p>
    <p>From what I remember, when the existing TTF composite
      specification was extended to support variable fonts the variable
      parameters were limited to the positions of the components and the
      side bearing points. That's both a more limited use of variable
      parameters than in the variable composite specification, and also
      a design somewhat determined by the way that ordinary glyphs work
      in the glyf table (e.g. GVAR is already the mechanism for making
      ordinary glyphs variable so why do something different?)</p>
    <p>When it comes to the use of the TVS or some alternative,
      therefore, one can't just look at the current spec and go "OK,
      that's more or less the same thing". Or at least I can't. <br>
    </p>
    <p>It sounds like you have already done a bunch of engineering to
      arrive at the current balance, but none of that information is
      part of the proposal, which leaves that part of it difficult to
      evaluate. What is the typical per-glyph overhead of the TVS
      headers, if any? How much benefit will the shared point numbers
      provide, if any? What is the overhead of a non-variable parameter
      compared with just specifying its value?</p>
    <p>The other typical way answering these questions -- going off and
      looking at a bunch of examples -- is hampered by the lack of
      shared prototype data. From what I recall there <i>is</i> test
      data that has been used to develop these proposals, but it's not
      public or shared with the group. <br>
    </p>
    <p>Perhaps one way out of this would be a brief "show the work"
      document on this specific subject, explaining why the proposed
      representation provides the best balance of file size and
      simplicity.</p>
    <p>Skef<br>
    </p>
    <div class="moz-cite-prefix">On 12/6/23 15:36, Behdad Esfahbod
      wrote:<br>
    </div>
    <blockquote type="cite"
cite="mid:CAF63+7UshXjBxC_K_VOvOVXhun89UXZ7xV6+sTAif4atX12Y-g@mail.gmail.com">
      <meta http-equiv="content-type" content="text/html; charset=UTF-8">
      <div dir="ltr">
        <div dir="ltr">Thanks Skef,
          <div><br>
          </div>
          <div>Comments below:</div>
        </div>
        <br>
        <div class="gmail_quote">
          <div dir="ltr" class="gmail_attr">On Wed, Dec 6, 2023 at
            4:09 PM Skef Iterum <<a href="mailto:skef@skef.org"
              moz-do-not-send="true" class="moz-txt-link-freetext">skef@skef.org</a>>
            wrote:<br>
          </div>
          <blockquote class="gmail_quote"
style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
            <div>
              <p>OK, the nitty-gritty:<br>
              </p>
              <p>The primary problem I see with the TVS is that --
                different from the coordinates in a glyph -- one would
                expect that in a typical composite glyph some parameters
                will be variable and some won't be. Positions, for
                example, will often be variable. Rotations and skews,
                probably not. Axis specifications -- it probably
                depends. </p>
              <p><i>Given the current spec</i> there are two means of
                dealing with non-variable parameters:</p>
              <ol>
                <li>You can treat them as pseudo-variable, essentially
                  adding 0 deltas for the index into the appropriate
                  contexts in the TVS</li>
              </ol>
            </div>
          </blockquote>
          <div><br>
          </div>
          <div>Note that zeroes are not encoded explicitly in the TVS. A
            flag and count is encoded of the number of zeroes.</div>
          <div><br>
          </div>
          <div><br>
          </div>
          <blockquote class="gmail_quote"
style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
            <div>
              <ol>
                <li>You can have each TVH sub-region "skip over" the
                  index.</li>
              </ol>
              <p>However, both of these solutions add overhead.</p>
              <p>In thinking about this more recently (i.e. after the
                discussion stopped) I've been wondering if it would just
                be best to add a mask to the VC entry, with one bit per
                parameter padded out to bytes, indicating whether the
                entry is variable or not. Then the TVS indexes can be
                reduced to only actually variable parameters, so that
                the bytes you pay for the mask are the only overhead.</p>
            </div>
          </blockquote>
          <div>That was indeed part of the original VARC proposal from
            Black Foundry. I changed that to only include a bit for
            "axis coordinates are variable", because I think the TVS
            does a good job of encoding zeroes.<br>
             </div>
          <blockquote class="gmail_quote"
style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
            <div>
              <p>That would deal with the <i>primary</i> problem.
                Whether that would leave the TVS as optimal, though,
                still seems like an open question.<br>
              </p>
            </div>
          </blockquote>
          <div><br>
          </div>
          <div>The IVS is inherently subpar for this purpose IMO.</div>
          <div><br>
          </div>
          <div>b</div>
          <div><br>
          </div>
          <div><br>
          </div>
          <div> </div>
          <blockquote class="gmail_quote"
style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">
            <div>
              <p> </p>
              <div>On 12/6/23 14:16, Behdad Esfahbod wrote:<br>
              </div>
              <blockquote type="cite">
                <div dir="ltr">Hi everyone,
                  <div><br>
                  </div>
                  <div>I support an external VARC table as well. Where
                    we got stuck is in the discussions of:</div>
                  <div><br>
                  </div>
                  <div>  <a
href="https://github.com/harfbuzz/boring-expansion-spec/issues/103"
                      target="_blank" moz-do-not-send="true"
                      class="moz-txt-link-freetext">https://github.com/harfbuzz/boring-expansion-spec/issues/103</a></div>
                  <div><br>
                  </div>
                  <div>where I was happy to go ahead and prototype a
                    table with TupleVariationStore. But Skef wants both
                    TupleVariationStore and ItemVariationStore paths to
                    be explored, and that was outside of my time budget.</div>
                  <div><br>
                  </div>
                  <div>It's good that we now have it on the agenda who
                    is supposed to work on it.</div>
                  <div><br>
                  </div>
                  <div>Thanks,</div>
                  <div><br clear="all">
                    <div>
                      <div dir="ltr" class="gmail_signature">behdad<br>
                        <a href="http://behdad.org/" target="_blank"
                          moz-do-not-send="true"
                          class="moz-txt-link-freetext">http://behdad.org/</a></div>
                    </div>
                    <br>
                  </div>
                </div>
                <br>
                <div class="gmail_quote">
                  <div dir="ltr" class="gmail_attr">On Wed, Dec 6, 2023
                    at 3:13 PM Liam R. E. Quin <<a
                      href="mailto:liam@fromoldbooks.org"
                      target="_blank" moz-do-not-send="true"
                      class="moz-txt-link-freetext">liam@fromoldbooks.org</a>>
                    wrote:<br>
                  </div>
                  <blockquote class="gmail_quote"
style="margin:0px 0px 0px 0.8ex;border-left:1px solid rgb(204,204,204);padding-left:1ex">On
                    Tue, 2023-12-05 at 04:14 -0800, Skef Iterum wrote:<br>
                    <br>
                    > Of course, in practice this means that we're
                    asking that variable<br>
                    > composites be taken out of the upcoming
                    proposal (and that if it<br>
                    > isn't Adobe will vote not to approve it, and
                    encourage others to do<br>
                    > the same). However, we want to stress that this
                    does not necessarily<br>
                    > mean we will not vote in favor later if further
                    research indicates<br>
                    > that glyf is the better way to go.<br>
                    <br>
                    Can we adopt a slightly different approach? We’re
                    looking at coming up<br>
                    with a proposal for an external table, outside GLYF,
                    as you/Adobe<br>
                    proposed, but in the meantime, since the ad hoc
                    meeting is on Monday, i<br>
                    can't really change the proposal we've submitted.<br>
                    <br>
                    However, we do see the motivation, and the document
                    is a working draft,<br>
                    so it can be changed, and that's fine.<br>
                    <br>
                    > Accordingly, we also suggest that how to go
                    about that research and<br>
                    > development, including who needs to be
                    involved, should be one topic<br>
                    > for the meeting next week. <br>
                    <br>
                    That's fine, i see Vlad added it to the agenda.<br>
                    <br>
                    And it'd be OK to vote for the existing Google
                    proposal to go ahead but<br>
                    with variable composites removed, of course. Or,
                    with the proviso that<br>
                    a proposal for an external table be developed at
                    least far enough for<br>
                    concrete discussion.<br>
                    <br>
                    I don't know that we can a new external-table
                    proposal ready by Monday,<br>
                    and in any case people won't have seen it. But we
                    can put it in the<br>
                    Boring Expansions repository and send email about
                    it.<br>
                    <br>
                    I'm sorry if we dropped the ball on the external
                    table proposal;<br>
                    silence in this case was not a sign of disagreement
                    or disapproval or<br>
                    anything; i should have pushed for discussion about
                    it internally<br>
                    here).<br>
                    <br>
                    Anyway, either way is fine, but i want to avoid the
                    situation where we<br>
                    end up with no variable composites at all by April,
                    so having at least<br>
                    one approach, albeit a flawed one, in the working
                    draft, might be<br>
                    better than none? What do you think? Again, it's a
                    working draft, so we<br>
                    can take things out if a better approach is chosen,
                    and that's true for<br>
                    any part of the proposals.<br>
                    <br>
                    Thanks,<br>
                    <br>
                    liam<br>
                    <br>
                    -- <br>
                    Liam Quin, <a
                      href="https://www.delightfulcomputing.com/"
                      rel="noreferrer" target="_blank"
                      moz-do-not-send="true"
                      class="moz-txt-link-freetext">https://www.delightfulcomputing.com/</a><br>
                    Available for XML/Document/Information
                    Architecture/XSLT/<br>
                    XSL/XQuery/Web/Text Processing/A11Y training, work
                    & consulting.<br>
                    Barefoot Web-slave, antique illustrations:  <a
                      href="http://www.fromoldbooks.org"
                      rel="noreferrer" target="_blank"
                      moz-do-not-send="true"
                      class="moz-txt-link-freetext">http://www.fromoldbooks.org</a><br>
                    _______________________________________________<br>
                    mpeg-otspec mailing list<br>
                    <a href="mailto:mpeg-otspec@lists.aau.at"
                      target="_blank" moz-do-not-send="true"
                      class="moz-txt-link-freetext">mpeg-otspec@lists.aau.at</a><br>
                    <a
href="https://lists.aau.at/mailman/listinfo/mpeg-otspec"
                      rel="noreferrer" target="_blank"
                      moz-do-not-send="true"
                      class="moz-txt-link-freetext">https://lists.aau.at/mailman/listinfo/mpeg-otspec</a><br>
                  </blockquote>
                </div>
              </blockquote>
            </div>
          </blockquote>
        </div>
      </div>
    </blockquote>
  </body>
</html>