On 15/11/2005, at 7pm, Benjamin BAYART wrote:

> Le Tue, Nov 15, 2005 at 10:51:30AM +1030, Will Robertson:
>>
>> So we get ... (other "1") (cseq "\previouslet") (cseq "\test")
>> and end up with the equivalent of \let\test=1 .
>>
>> Ohhh, but I think I get it now. This is no more useful than actually
>> looking at nodes, since anything that survives to be \previouslet
>> will be turned into a node anyway.
>
> Yes. It's more clearly seen with:
>
> \hbox{a}\previouslet\test
>
> What is \test? The right brace? The box? The box is not a token any
> more. So we have to keep on that:
> - look ahead for tokens
> - look backward for nodes.

Yes, it is [or would be] the right brace. That is the previous token,  
before what came prior to \previouslet triggered the building of a  
node. I'm not saying this would be a particularly useful feature at  
this point :)

I've been convinced that \previouslet isn't useful, but not that it  
isn't possible; but who wants possible if it's useless, right?

Thanks all for the exposition,

Will