Record feedback to UTC#147 feedback about CHAI

[srl295]

From http://www.unicode.org/L2/L2016/16123-pubrev.html#Error_Reports :

Date/Time: Fri May 6 07:33:00 CDT 2016
Name: William Overington
Report Type: Feedback on an Encoding Proposal (L2/16-105)
Opt Subject: Feedback on Coded Hashes of Arbitrary Images proposal

I write to comment on the following document.

http://www.unicode.org/L2/L2016/16105-unicode-image-hash.pdf 

I support the proposal.

I write to make two suggestions please.

1. In section 2.5 of the pdf document, at the end of the section is the following.

> > Depending on the protocol, as the input arrives, the receiver may have
> > some ambiguity about when the sequence of CHAI characters ends. A receiver
> > may choose to wait until the next non-combining character (signaling the
> > end of the combining character sequence), or a protocol-defined end-of-
> > message signal, before retrieving the emoji description.

I opine that it would be better to encode

U+EFFFB IMAGE HASH CODE SUBSET COMPLETED

and to use that character at the end of a sequence of IMAGE HASH characters so
that when that character is reached, a definite indication that the hash code
subset has been completed is received from within the Unicode plain text
message.

2. In section 2.5 of the pdf document there is the following.

> > Otherwise, the receiver displays the base character while it attempts to
> > retrieve an emoji description whose hash matches the encoded hash prefix.

As the method is so that arbitrary images can be referenced from a plain text
sequence, it seems that no existing base character may be suitable for every
arbitrary image.

So, could a

BASE CHARACTER FOR AN ARBITRARY IMAGE

be encoded please,

either as U+EFFFA or, as it is a displayed character, with a code point in plane 1.

It is possible that U+1F5BC FRAME WITH PICTURE could be used, yet I opine that
a specific BASE CHARACTER FOR AN ARBITRARY IMAGE character becoming encoded
would be a better solution as that would clearly indicate that an arbitrary
image not necessarily based upon any character in regular Unicode is being
referenced.

William Overington

Friday 6 May 2016

[srl295]

U+EFFFB IMAGE HASH CODE SUBSET COMPLETED

I think we don't need to mark when the hash is completed, the rendering should degrade gracefully if the hash is shorter than expected.

BASE CHARACTER FOR AN ARBITRARY IMAGE

The fallback behavior was not as well received, so yes, using such a base char could be a good way forward.

[srl295]

closing.