Am Fr., 30. Sept. 2022 um 22:35 Uhr schrieb John Cowan <[email protected]>:

[...]


>  As I
>> said, not using symbols in SRFI 233 is only a solution as long as the
>> programmer doesn't use string->symbol, etc.
>>
>
> Daphne convinced me that this concern is pointless, because SXML and
> various JSON libraries are already interning untrusted input right and
> left.   So section names and keys (but not values) will be symbols.  In
> addition, lines without brackets or = will be values with key #f instead of
> keys with value "" (but this change has not yet been made).
>

Okay. In any case, adding a warning as a note to SRFI 233 may help to make
this problem better known.


> Arvydas, can you make the above fix to johnwcowan/srfi-233?  I don't quite
> follow the logic of the parser.
>
> I don't understand why this should be an oddball use case. And why
>> "jumping through hoops"? tt would be enough to say something with the
>> effect that the generator and the accumulator are not lazy.
>>
>
> Well, no.  To impose this restriction, the SRFI would have to say that the
> generator must be lazy in the sense that it must not consume from its port
> more than necessary, and the accumulator must be eager in the sense that it
> must push to its port all that it has available when it is called.
>

You are right, of course.  This is also the behavior of `read` and `write`
in Scheme.

Reply via email to