On 25/02/2009, at 6:38 AM, Joseph Wright wrote:

> 1) Are there any other missing primitives?  The other \every...
> primitives should (I think) not be used outside of the kernel, and so
> don't need to be worried about.

Hmmmm, I actually think that we *should* provide access to these via a  
well-defined mechanism. That might be an xpackage job, or maybe they'd  
fit into \toks. But I'd hate to have to give them up (the reason  
they're banned in LaTeX is just that they get overwritten, right? At  
least in the case of \everypar).

> 2) How should the missing three (plus any others) be named.  The  
> current
> idea is, I think, something like:
>
> \endlinechar	\l_char_endline_int
> \evereof	\l_ior_eof_toks
> \scantokens	\tlist_rescan:n
>
> Is this logical: are there any other suggestions?

Except for the other \every's, I'm not aware of anything else missing  
-- but I haven't manually checked through the listings in l3names.  
There are also a bunch of pdfTeX primitives we're not using.

> 3) Do the three primitives need to be accessible?  I think it's hard  
> to
> argue against \scantokens, at least.

Yeah, I think we need them all.

Will