Ftp-WG: NLST vs MLSD

Jeffrey Altman writes:
> when kiddies get around to implementing MLSD

You are continuing to miss the point. The clients you're talking about
are broken. They are handling wildcards in a way that

   * is not guaranteed to work by the FTP protocol and
   * _does not work_ with some deployed protocol-compliant servers.

Those clients should be fixed. I've explained the correct way to handle
wildcards. The correct way works with _all_ FTP servers. One example of
a working client is ftpfs.

> Netscape

I'm curious: Since when did Netscape have multiple-file FTP retrieval?

> It is also wrong to assume that bandwidth to
> the endpoints can handle large quantities of raw unneeded data.

You are talking about a ridiculously minor bandwidth issue. We could
save more of the Internet's bandwidth by removing the ``P'' on the first
line of every HTTP connection. Don't you have any real work to do?

> To say that FTP should not be used for this class of application is absurd.  

FTP should not be used for this class of application. The protocol does
not provide disk commitment, for example, so it can't be used for
reliable transaction processing.

If you aren't using FTP, but rather a private extension to FTP with some
transaction features added, you're an incompetent protocol designer.
Starting from FTP as the base for a new protocol is like starting from
OS/360 as the base for a new operating system.

---D. J. Bernstein, Associate Professor, Department of Mathematics,
Statistics, and Computer Science, University of Illinois at Chicago