From: David Starner (
Date: Mon Oct 24 2005 - 14:59:41 CST

  • Next message: Johannes Bergerhausen: "Re: LAO LETTER FO SUNG... wikipedia"

    On 10/24/2005 5:53 AM, Philippe Verdy wrote:

    > Would you really do that so much blindly? This would mean importing
    > tend of thousands of identifiers in your program, which would cause
    > much more conflicts. If they are imported in a namespace, this will
    > clutter your program for nothing, with excessive and unneeded
    > compilation time due to the creation of management of large symbol
    > tables filled by your generated names and the true identifiers used by
    > your application or by imported system identifiers.

    Only one programming language commonly in use doesn't do namespaces,
    so the conflict of names with other names is not a concern. Just stdio
    on some systems can drag in 20,000 lines of headers, and I would
    expect simple constants to be quicker to process than what's in those
    20,000 lines.

    > Really you will define carefully only some identifiers for characters
    > you need to reference specifically in your program, and the name will
    > you choose for those identifiers has no link to the ISO/IEC 10646
    > normative character name. You will define its value to the codepoint.
    > So you don't need those character names. Really.

    No, actually, my programming language (Ada) defines identifiers for
    all the Latin-1 characters, and I see no reason not to define standard
    identifiers for all the Unicode letters. It reduces the number of bugs
    where programmers disagree about which character
    Unicode.Quotation_Mark is, or where programmers define
    Latin_Capital_Letter_N_With_Long_Right_Leg to be 16#0202#. Standard
    libraries are better tested, and when they have bugs, they only need
    to be fixed once.

    This archive was generated by hypermail 2.1.5 : Mon Oct 24 2005 - 15:00:51 CST