Error Recovery (was: Re: Should we say anything on security?)

On 09/12/2012 10:45 AM, Uche Ogbuji wrote:
> On Wed, Sep 12, 2012 at 7:56 AM, Mike Sokolov <sokolov@falutin.net 
> <mailto:sokolov@falutin.net>> wrote:
>
>     On 09/12/2012 09:49 AM, Uche Ogbuji wrote:
>>     On Wed, Sep 12, 2012 at 7:41 AM, Mike Sokolov
>>     <sokolov@falutin.net <mailto:sokolov@falutin.net>> wrote:
>>
>>         On 09/12/2012 09:33 AM, Uche Ogbuji wrote:
>>
>>             I think this is the way almost every implementor has
>>             interpreted it as well.  Some, such as libxml will take
>>             advantage of the "in a normal way" clause to at least try
>>             to show the user any further fatal errors beyond the
>>             first, to make fix-up a bit less painful, but yeah that
>>             hardly counts as liberal acceptance, and anyway most
>>             parsers do stop dead at the first fatal error.
>>
>>
>>         Not completely - MarkLogic for example, provides "fixup"
>>         capabilities in its parser that include undeclared ISO xml
>>         entity handling, and I think will even fix well-formedness
>>         problems a-la tidy.
>>
>>
>>     Err, that is not an XML parser, then, any more than Tidy is, or
>>     html5lib, though both can parse XML-like thingies.
>>
>     Isn't your argument circular?  "No parser implements fixup,
>     because one did, it wouldn't be a parser"
>
>
> Err no.  That's not remotely what I said.  I think I've been pretty 
> clear in what I've said, which you quoted above.
>
I said "MarkLogic provides fixup capabilities"

You said "that is not an XML parser, then", which I read as "Anything 
that provides fixup capabilities is (ipso facto) not an XML parser".

At least I think that's what you meant?

What a parser is allowed to do when presented with certain kinds of 
errors, and still call itself an XML parser is something that I've never 
understood well, so I don't really want to enter into a prolonged 
discussion about it - I'm sorry if I misconstrued your statement.

Most XML processors (I won't say parser, to avoid continued 
misunderstanding I hope) seem to barf at the first sign of trouble, and 
that's true.  I think we agree on that, at least?

More interestingly for this discussion, the MicroXML position on error 
recovery *does* allow reporting a data model in all kinds of cases so 
long as it is accompanied (how?) by a statement that the document is not 
conforming.  I'm glad about this.

-Mike