[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: CLtl Natural Languages Subcommittee



    Date: 11 Aug 87 15:16 PDT
    From: masinter.pa@Xerox.COM
    I had hoped that this committee could deal also with the issue of font
    and font attributes as well as character codes. 
Yup.  I plan to propose we delete 'em.  Details later.
						    I'd been filing things
    under the heading of "Common Lisp Characters", since that seemed to have
    the broader charter.  

    My comments at the last X3J13 committee is that, while there may still
    be important reasons for retaining a user-visible distinction between
    thin-simple-string and simple-string, there seemed to be little or no
    reason to have any visible distinction between thin-string and string,
    since the general string case, with displacement and the like, can be
    implemented as efficiently.

This is a lot easier for you or I to say, on our special Lisp engines,
than it is for those on "stock" hardware.  But even for us, it's not
really true.  Creating a large string of the wrong size, and then copying
the whole thing when a "fat" character comes along, could prove quite
expensive.  And on our system, once that was done, every reference to that
string would be slowed down by an extra memory reference.  Everybody else
would have to pay this price ALL the time, except when they can use simple
strings.

Don't forget that (AND STRING (NOT SIMPLE-STRING)) does NOT mean that
the string is displaced.  It may be a non-displaced non-adjustable
string with a fill-pointer.  What you're proposing really means changing
this so that, on stock architectures, (AND STRING (NOT SIMPLE-STRING))
implies that it's displaced, so that the data can be "fattened".

    This modification of the JEIDA proposal removes most of its complexity
    while retaining most of its benefits.

I don't think it removes the complexity; it introduces complexities of
its own.  Of course, it also has benefits of its own.  But I don't want
to waste time designing it unless we have some assurance that it really
isn't going to be a burden for the stock architectures, and the understanding
I have from the conversations I've had is that it would be a burden for them.

I'm about to leave until Labor Day.  I'll have lots more stuff to send when
I get back.