W3C home > Mailing lists > Public > public-ixml@w3.org > March 2021

Minutes ixml call 2020-08-27

From: Steven Pemberton <steven.pemberton@cwi.nl>
Date: Wed, 17 Mar 2021 15:07:39 +0000
Message-Id: <1615993544553.4064973254.2768051427@cwi.nl>
To: public-ixml@w3.org
ixml call 27 Aug 2020
Present: CMSMQ, TH, SP

AGENDA

Specification
Implementations
Test Suite
AOB
ACTIONS

ACTION: Steven to forbid "." at end of name [wihdrawn]
ACTION: Steven to specify what happens when a name isn't an XML name
ACTION: CMSMQ to send augmented grammar
ACTION: Steven create Github repo

Next call: Thursday 2020-09-24 14:00Z
https://meet.google.com/dfz-rwpj-opq

SPECIFICATION

M: Sanity check.
... version I am working from is dated January
S: 23 July has corrections from you
M: A practical instance of ambiguity.
a: b, c, d, e.
... a= b c d e.
.. failed because of nonterminal ending with "."
... you Steven suggested forbidding a name ending with a dot
S: I did.
ACTION: Steven to forbid "." at end of name

T: Should we restrict rule names to XML names?
S: Not all rules are serialised.
M: what if you do try to serialise a name that isn't an XML name
S: Error
T: Recoverable error
ACTION: Steven to specify what happens when a name isn't an XML name

M: What works well in XSLT etc is that from SQL people the WG adopted 
implementation defined and implementation dependent
... defined is where the implementation defines what happens, at same level 
of specifity as spec;
... implementation dependent - no obligation to document more than 
"something will happen".
... so the recovery should be implementation defined, not dependent
T: Sure. Minimum an error.
S: Do we really want to recover?
M: More convenient for the user? Required failure works for me.
T: Fine either way.
M: One way full stops is not what Tom said. XML names *can* end with a 
fullstop.
... otherwise we would have to demand whitespace before the terminal full 
stop, which would look horrible.
S: It could mainly be a problem with LL1
M: I am parsing LL1 (recursive descent).
... mark and tmark are non LL1
... and 'member' has to be rewriten
... it may be just a problem with my parser.
T: Do we make the assumption in the spec explicit?
S: You can bootstrap, and then the only problem is the initial 
bootstrapping parser.
M: I will withdraw the suggestion of restricting fullstop at the end of a 
name, and mention it as a requirement of my parser.
M: I will include it in the testsuite
T: I would like a link to testsuite in the spec
... references in the spec as it describes feature.
M: Maybe multiple views/trails to unclutter the spec.
T: CSS and/or XForms
M: Let's come back to this when we have a testsuite.
M: I will send my augmented grammar
ACTION: CMSMQ to send augmented grammar

IMPLEMENTATIONS

M: I have finished my rec desc parser. Not yet tested in great detail.
... generates an XML rep of the grammar
... an ixml processor restricted to a single grammar
... written in XSLT

TEST SUITE

T: Alas, no opportunity to work on this since Balisage
... I want to add a field for ref numbers.
... a unique ID
... I'm using cucumber/gherkin
M: If I use Gherkin, I could convert it to XSpec or shell.
T: Sure
... my goal is to write a grammar for gherkin, and parse to XSpec
S: How about using ixml?

https://github.com/eXpertML/JayParser/blob/master/src/test/resources/gherkin/ixml.feature

T: that's my starter
... self describing; I don;t like that the table of things to check can't 
contain grammars or XML fragments - no line breaks allowed
M: Let's try Gherkin and see what it gives us
... I like the idea of having a non-XML syntax.
T: There's no such thing anymore!
[Laughter]
S: Where are we going to put all this stuff?
M: Do we want to create a repo for the three of us, or develop now and 
merge.
T: I'd like to have a test suite before I finish the parser
... I'd like the two of them together in a repository
... for instance using Git
... and then we are also ready for contributors
... they should live together indenpendent from the implementations.
M: Gitlab or Github?
T: Little difference
... Github is a bit friendlier.
S: Let's use Github
T: We need some starting point for the tests. We neeed test instances.
... I have attempted to describe the format in the Gherkin
S: I'm happy putting the spec there
T: Initial action then
ACTION: Steven create Github repo

T: Who is owner? Individual or org.
... On Github you can create an org.
Steven: I'll research and do it.
T: And put minutes there.

AOB

M: Do we want to testsuite as an onion?
... start minimal and build
S: Sounds good.
... good for implementations too

Next call: Thursday 2020-09-24 14:00Z
Received on Wednesday, 17 March 2021 15:07:54 UTC

This archive was generated by hypermail 2.4.0 : Wednesday, 17 March 2021 15:07:55 UTC