On 12/26/2025 2:11 AM, [email protected] via Unicode wrote:
From https://vt100.net/charsets/technical.html it appears that a
possibly intended grid is the following:
There's an image in that file that shows this relation but I didn't
notice that on first reading (the layout is really funky on a wide screen):
(green on black image from https://vt100.net/charsets/technical.html)
And this corresponds to your analysis of the 5-story summation sign.
(Leaving out the middle column and any row that is only used for the
middle column would yield the 3-story version.)
03/01 02/03 03/05
03/03
03/07
03/04
03/02 02/03 03/06
From the image of the character set it is clear that 03/01 occupy the
right half of a cell, and 03/05, 03/06 and 03/07 occupy the left half of
the cell, which is as expected if they are supposed to be right-joining
and left-joining. This underscores analyzing them as a coherent subset,
with 03/02 shared between summation and square root.
(character set table from https://vt100.net/charsets/technical.html)
The use of 03/03 and 03/05 is possibly not constrained to the summation
sign (e.g. could be used for some kind of diagonals) but the glyph
design is constrained to match up the the other characters in this subset.
Because we already have glyph pieces for large operators and fences, I
don't see why there should be a precedent that prevents us from having
mappings that work. However, when making the case, please provide the
images in the proposal document, not just hidden in a reference.
It's useless to discuss this stuff with words alone.
A./
And by checking the proposals to see how those characters were
intended to be mapped to Unicode:
03/01, 03/02: L2/98-354 maps them to private use E0AD and E0AC, then
L2/00-159 remaps to STIX summation symbol halves U+23B2 and U+23B3
respectively (the proposals do not acknowledge any usage of other
portions of the summation symbol)
02/03: L2/00-159 mentions it as U+2500 or U+23AF
03/03, 03/04: not mentioned in the proposals, but
www.columbia.edu/kermit/ftp/charsets/dectech.txt
<http://www.columbia.edu/kermit/ftp/charsets/dectech.txt> names them
as 'Upper left to lower right diagonal line' and 'Upper right to lower
left diagonal line', most likely referring to existing U+2572 and
U+2571 respectively
03/05, 03/06: L2/98-354 maps them to private use E0AE and E0AF,
then L2/00-159 remaps to existing ceiling/floor symbols U+2309 and
U+230B respectively
03/07: not mentioned in the proposals, but
www.columbia.edu/kermit/ftp/charsets/dectech.txt
<http://www.columbia.edu/kermit/ftp/charsets/dectech.txt> names it as
'Ket (Large right angle bracket)', most likely referring to either
U+232A or U+3009 (as U+27E9 did not exist at the time)
Dnia 26 grudnia 2025 09:41 Asmus Freytag via Unicode
<[email protected]> napisał(a):
From a quick glance, it looks like Unicode so far has one fully
formed glyph piece for top and bottom each for the summation
symbol. They are enocoded at U+23B2 and U+t23B3. These are
intended to allow a two-story sigma, but not an infinitely large
one (unless you use an OpenType font that scales stroke width down
in relation to height for larger font sizes).
The Dec Character set appears instead to have the left / right
corners for the top and bottom of the summation symbol encoded
separately which would allow a wider top or bottom. There's also a
piece that looks like a shallow right-pointing V, which sets this
up for a three-story symbol.
If the DEC Technical top left and bottom left corners (03/01 and
03/02) were used for a 2-line summation symbol, then the centering of
the horizontal strokes has the effect of making the overall 2-line
summation have the black-body height of the sum of line height and
stroke weight, and the additional diagonal connections that cause the
middle corner to end up on the right edge, which could be considered a
font-level variation where the summation halves also happen to be
usable for larger connections, similarly to how arrow characters may
or may not have to connect to bounding box edges depending on the
compatibility scenario.
The 'shallow right-pointing V' (which is 03/07) has diagonal
connections that also occur in the Ohio Scientific character set,
resulting in it being proposed in L2/21-235 and incorporated in
Unicode 16.0 as U+1FBDB BOX DRAWINGS LIGHT DIAGONAL UPPER LEFT TO
MIDDLE CENTRE TO LOWER LEFT.
Additional diagonal pieces seem to have the correct angle to allow
both a 3 story and a 5-story summation symbol. This of course,
would be wider than a character cell, so the top and bottom are
broken apart, so they can be wider in a monospaced font.
That the horizontal lines with the down/up tick at the right end
are supposed to extend to top/bottom bar on the sigma, with the
ticks being the "serifs" on the Sigma may not have registered
because they simply look like rotated floor / ceiling style symbols.
I don't think there is any rotation involved (it wouldn't make sense
given the character grid). The proposals L2/98-354 and L2/00-159 do
not acknowledge usage of other characters than 03/01 and 03/02 to form
the summation symbol, so the identification of 03/05, 03/06 as
ceiling/floor symbols could be a result of incomplete information on
the usage of the characters.
So the summation symbol pieces would seem to be set up out of the
box for a five story design that is two cells wide. What is
encoded in the 2300 block is two stories tall and one cell width
(unless Opentype tricks are used to make the pieces larger without
changing the stroke width, but if you can do that, you wouldn't
need pieces in the first place).
The existing characters cannot be used for a design that uses
glyph pieces to give a five story x 2 cells design.
Due to the diagonal connections there would have to be at least 3
columns for a 5-row summation symbol.
It would help if any proposal documents actually showed the
assembled symbols / preferably in the context of a sample formula.
Showing just the character set is not as helpful.
And it would be useful to show an attempt to do the same with the
characters already encoded in the standard to show the difference
in design of hte glyph piece set for 2-story vs. 3 or 5 story designs.
A./
Such a proposal would also have to assess potential candidates for
unification:
— U+2309 and U+230B: this is what 03/05 and 03/06 were identified as
in L2/00-159, but the typical usage of bracket sized symbols is
completely different from the corners of summation symbol and likely a
mutually exclusive usage
— U+231D and U+231F: I'm not sure exactly what the usage of those
corner characters is, or whether centering of horizontal stroke and
horizontal connections to other characters is allowed
— U+2510 and U+2518: those definitely cannot be used, as they are used
in DEC Special Graphic Character Set and connect to bottom or top
edge, having a distinct appearance (and in
http://bitsavers.org/pdf/dec/terminal/vt340/EK-VT3XX-TP-001_VT330_VT340_Text_Programming_Mar87.pdf it
is mentioned that DEC Special Graphic and DEC Technical sets may
replace GL or GR sets, implying the possibility of the two coexisting
in the same document if one is used in GL and the other in GR)
— U+1CC1C and U+1CC1B: those are Sharp MZ compatibility characters,
but connect on both left and right and to bottom and top edge, so it
is unlikely to be usable for summation corners
