- From: Ian B. Jacobs <ij@w3.org>
- Date: Sun, 28 Dec 1997 17:12:48 -0500
- To: "Adam M. Costello" <amc@cs.berkeley.edu>
- CC: www-html-editor@w3.org, w3c-html-wg@w3.org
Adam M. Costello wrote: > > Many of the comments point out typos. Some point out confusing, > misleading, or imprecise parts of the spec, and suggest > clarifications or additions (unless I was baffled). > > Sorry I didn't look at the spec when it was still a Proposed > Recommendation, but the semester just ended. Adam, Thank you for your careful reading of the spec - where were you during the final editing process! I don't believe the W3C process will allow us to release a revision of the HTML 4.0 Recommendation to correct typos and small errors. Therefore, we won't be able to take your comments into account unless we prepare a more substantial revision of the language. Until then, we will be maintaining the errata sheet, which I have updated based on your comments. I'm forwarding your more substantial remarks (and my comments) to the Working Group for perusal. Thanks again for your good work! -- Ian > 8.2.4 Overriding the bidirectional algorithm: the BDO element > > One reason for this may be that the MIME standard ([RFC2045], > [RFC1556]) favors visual order, i.e., that right-to-left > character sequences are inserted right-to-left in the byte > stream. > > I don't think this means what was intended. My best-effort > interpretation of "right-to-left character sequences are inserted > right-to-left in the byte stream" is that the rightmost character > appears first in the byte stream. But that is the opposite of RFC > 1556 visual directionality, which requires the leftmost character > to appear first in the byte stream. I strongly recommend using > phrases like "leftmost character first" and avoiding phrases like > "right-to-left in the byte stream", because byte streams do not have > a left and right, only an earlier and later. I will look into this issue. > 8.2.5 Character references for directionality and joining control > > Mirrored character glyphs. In general, the bidirectional > algorithm does not mirror character glyphs but leaves them > unaffected. An exception are characters such as parentheses (see > [UNICODE], table 4-7). > > Although the Unicode character names and example glyphs are > available online, the text of the spec is not, so I wish the HTML > spec would elaborate a bit on the mirroring of parentheses. If > characters the characters ( and ) were called "open parenthesis" and > "close parenthesis", I could understand why their appearance would > depend on the directionality of the text. But they're called "left > parenthesis" and "right parenthesis", so I don't see why they would > ever be mirrored. In right-to-left text, you would obviously begin > a parenthetical with a right parenthesis, and end it with a left > parenthesis, correct? I will also look into this. > 9.1 White space > > authors should not rely on user agents to render white space > immediately after a start tag or immediately before an end tag. > > What about the converse? Should authors also not rely on user > agents *not* to render whitespace immediately after a start tag? > For example, may authors assume that these will be rendered the > same: > > <li>foo > <li> foo > > or should authors always use the first form? It is my opinion that the sentence in the spec subsumes the converse. > > 9.3.4 Preformatted text: The PRE element > > width = number [CN] > This attribute provides a hint to visual user agents about > the desired width of the formatted block. > > By definition, preformatted text already has a width, which can be > determined by scanning it and noticing the length of the longest > line. Maybe you mean that this attribute provides a hint, not about > the width of the text, but about the width of the window for which > the text was formatted. I believe the hint is for user agents so that they may reserve space before formatting. > > When handling preformatted text, visual user agents: > May leave white space intact. > May render text with a fixed-pitch font. > May disable automatic word wrap. > > Shouldn't each "may" be "should"? Authors usually depend on these > for vertical alignment. Since these are formatting issues, and user agents vary significantly, we adopted less restrictive wording. > 11.4.2 Categorizing cells > > In order to determine, for example, the costs of meals on 25 > August, the user agent must know which table cells refer to > "Meals" (all of them) > ^^^^^^^^^^^^^^^^^^^^^ > > No, only cells in the Meals column refer to meals. Maybe you meant > "which table cells refer to "Expenses" (specifically, Meals)". No, I believe "Meals" (all of them) is correct, although perhaps somewhat vague. > 13.3.4 Object declarations and instantiations > > <P><OBJECT declare id="tribune" ... > <PARAM name="font" valuetype="object" value="#tribune"> > > Is the pound sign supposed to be there? Section 13.3.2 said: > > object: The value specified by value is an identifier that > refers to an OBJECT declaration in the same document. The > identifier must be the value of the id attribute set for the > declared OBJECT element. > > That suggests to me that the PARAM element should have > value="tribune", with no pound sign. I believe you're correct, but I have not yet added this to the errata sheet. > 13.6.1 Client-side image maps: the MAP and AREA elements > > usemap = uri [CT] > This attribute associates an image map with an element. The > image map is defined by a MAP element. The value of usemap > must match the value of the name attribute of the associated > MAP element. > > Since the value of the usemap attribute is a URI, it should be > permissible to refer to a MAP element from another document. None > of the examples do this. Is it allowed? If the specification doesn't explicitly forbid it, it is possible. I believe the Working Group chose to underplay this feature of the language. > 17.3 The FORM element > > The value is a space- and/or comma-delimited list of charset > values. > > This attribute specifies a comma-separated list of content types > > Throughout the spec, some attribute values are space-separated, some > are comma-separated, and some are space- and/or comma-separated. Is > there a simple rule that one can memorize, rather than consulting > the spec every time? If so, this rule should be stated somewhere. Unfortunately, there is no simple rule. For historical reasons, some lists are space-separated while others are comma-separated. > 17.11.2 Access keys > > accesskey = character [CN] > > How is this case neutral? Doesn't it have to be either case > sensitive or case insensitive? Am I allowed to have one control > with an accesskey of "C" and another with an access key of "c"? (I > vote no.) A "character" refers to a single character of the document character set. Since "c" and "C" differ in the document character set, there is no case involved. > By the way, shouldn't the spec say that no two controls in the same > document should have the same accesskey? It probably should. > On forms: > The examples use 'Content-Disposition: attachment' in the subparts, > rather than 'Content-Disposition: file'. Are both correct? Is one > preferred? This material has been taken from RFC 1867. If the RFC doesn't specify this, the HTML 4.0 Recommendation won't either. > > 24.4 Character entity references for markup-significant and > internationalization characters > > Entities have also been added for the remaining characters > occurring in CP-1252 which do not occur in the HTMLlat1 or > HTMLsymbol entity sets. These all occur in the 128 to 159 range > within the cp-1252 charset. > > What is CP-1252? It doesn't seem to be defined or referenced > anywhere. Also, either capitalize the second occurrence or > decapitalize the first. Good question. I hadn't noticed this before.
Received on Sunday, 28 December 1997 17:13:29 UTC