But we're not yet ready to send such an article back to a client. The
reason is most likely the fact that we read the entire article into
memory in fetch-article. The new sensible idea is to not read the
body of articles into memory, but delay the body reading until the
last moment---of sending it out to the client.
Nevertheless, we're getting needlessly ambitious here because we don't
intend to support file attachments anyway. We might do it just as an
exercise, if the end result looks good. I think it's a good exercise
and I think we should do it.
It also includes scripts for formating Lisp code for people who will
read Lisp code directly. It doesn't cost us much to organize the Lisp
output a bit. In a similar spirit, I'm not including NOWEB as part of
the build process. (Most users will not have NOWEB around to use it.)
So while the Makefile is pretty short because there's little to do,
Anyfile is the ANYWEB makefile. ANYWEB is my modified, personal
version of NOWEB for Windows. (The way I use this second makefile is
to make a shell alias called /amake/ that automatically invokes ``make
-f Anyfile'' for me.)
Notice that if a user doesn't change his password, we'll keep using
sxhash forever, which I think is quite okay. The problem with sxhash
is not only that it's not cryptographic, but different systems will
produce different hashes. For instance, Windows will produce one hash
and FreeBSD will produce another. We're better off with something
more deterministic such as SHA256.
- Converts Makefile to UNIX line termination.
Makefiles cannot be formatted with DOS CRLF because, otherwise, we
could not escape \n to continue on a second line. We end up escaping
\r and not \n.
- Fixes remove-account!
Procedure delete-if ``may modify sequence'', but we cannot be sure it
will modify it. There are cases in which it does and there are cases
in which it doesn't. Seeing it did modify in one case, I incorrectly
assumed it would modify in all cases---such is life. Since I do want
to modify it always, I wrote delete-if*, which in calls setf to be
sure the list is overwritten.
- Avoids (load'ing "~/.sbclrc") and, instead, does what Quicklisp does.
It's not always the case that ~/.sbclrc exists. But Quicklisp in each
installation knows what to do. So we do what Quicklisp does in each
script that we use.
- Frees us from using sb-ext:{file-exists,file-does-not-exist}
Turns out SBCL 1.2.4.debian doesn't have these symbols.
- Adds command-line parsing.
- Adds install target to Makefile.
- Rewrites Makefile with a new strategy.
- Adds the daemon-tls/ service example.
- Adds the daemon service example.
- Convers all text files to LF instead of CRLF.
