NEWS To-do list: Option to enable EMACS-like user folder versioning on each run. The IMAP support is naive. Chris Newman, one of the IMAP maintainers, criticized it as follows: ------------------------------- CUT HERE ----------------------------------- On Wed, 18 Sep 1996, Eric S. Raymond wrote: > 1. I do one one SELECT, at the beginning of the fetch. > > 2. I assume that I can pick an upper bound on message numbers from the EXISTS > reponse. Correct. > 3. If there is an UNSEEN nnn trailer on the OK response to SELECT, I assume > that the unseen messages have message numbers which are the nnn consecutive > integers up to and including the upper bound. > > 4. Otherwise, if the response included RECENT nnn, I assume that the unseen > messages have message numbers which are the nnn consecutive integers up to > and including the upper bound. These will only work if your client is the only client that accesses the INBOX. There is no requirement that the UNSEEN and RECENT messages are at the end of the folder in general. If you want to present all UNSEEN messages and flag all the messages you download as SEEN, you could do a SEARCH UNSEEN and just fetch those messages. However, the proper thing to do if you want to present the messages when disconnected from the server is to use UIDs. To do this, you remember the highest UID you have (you can initialize to 0), and fetch everything with a higher UID. Ideally, you shouldn't cause the SEEN flag to be set until the user has actually seen the message. This requires STORE +FLAGS SEEN for those messages which have been seen since the last update. The key thing to remember is that in IMAP the server holds the authoratative list of messages and the client just holds a cache. This is a very different model from POP. ------------------------------- CUT HERE ----------------------------------- IMAP extensions for secure challenge-response. Recode POP2 to use the same driver/method strategy as POP3/IMAP, so all three protocols will be able to forward messages through the generic driver to SMTP. (This requires that we find a POP2 server to test with.) 3.2 (Mon Sep 23 13:29:46 EDT 1996): * RPOP support (coded at a user's request but untested). * Ported to QNX (see the Makefile). * Add Michael Schwendt's code for improved sizeticker. * Improved RFC822 parsing (thanks to Rob Funk). * Move the per-user lockfile to /tmp so it gets cleared at reboot time. * Warn users that running concurrent instances of popclient is a bad idea. * Try USER and HOME to set defaults before going to the password file. This should work better in Sun NIS environments. 3.1: * MDA arguments are now dumped when using the -V option. * Sendmail delivery from background seems to work now. * We have IMAP2bis/IMAP4 support. * Code now autoprobes for a POP3, IMAP, or POP2 server if no protocol is specified. * SMTP forwarding support. Thanks to Harry Hochheiser for this simple but clever idea. It's now the default delivery mode. * If no UNIX From line is found, popclient will now synthesize a correct line from the RFC822 From line. * It is now possible to specify the host TCP/IP port number to connect to. 3.05: * Experimental support for RFC1725-compliant POP servers with the UIDL command and without LAST. 3.04: * Logfile option works. 3.03: * Minor bug fixes for password querying and redirection to stdout. 3.02: * Correct buggy processing of nokeep/noflush/fetchall. * Fix buggy -mda option processing. * Added -N/--norewrite option. * Delivery via sendmail now works in non-daemon mode. 3.01: * Fixed a lexical analyzer bug in quoted-string processing. * Fixed a bug in dump_options that caused username to be displayed incorrectly. * The lock assertion code was in the wrong place relative to the daemonize() call. 3.0: Eric S. Raymond hacked extensively on 3.0b6 and took over the package with the consent of Carl Harris, the original implementor. Some of the 3.0 feature additions were inspired by Sean Oh's fetchpop 1.8 code, and a few use code directly lifted from fetchpop. Here are my (Eric's) change notes: CONFIGURATION AND BUILDING * The autoconfigure script incorrectly assumed that all Linuxes use /usr/bin/deliver. Under Linux it now checks for both /usr/bin/delivermail and /bin/mail. * I added a distribution-maker production to Makefile.in. OPTIONS AND COMMAND LINE * I have removed the -p command-line option. Given that there's a .poprc facility there is no excuse for encouraging users to put plaintext passwords in scripts which might be readable. * Calling popclient with no arguments now causes it to query or operate on every host in the .poprc file. * I have made --version more useful by having it dump the computed connection options for each server specified. * The user can now explicitly set an MDA (such as procmail) with the new option -m or -mda. Various possible MDAs are listed on the man page. POPRC FILE SYNTAX * The .poprc lexer now supports "-enclosed strings which may contain whitespace. * I added a --yydebug option to enable .poprc parser debugging at runtime if the parser was generated with --debug. It's not documented. * You may now have a `defaults' entry in .poprc which sets overrideable values for other entries. See the man page for details. * It is now possible to set keep, flush and fetchall in your .poprc file. * Fixed incorrect numbering of source lines in .poprc parse error messages. * The configure.in specification no longer uses the obsolete AC_TRY_COMPILE macro (it uses AC_TRY_LINK instead). MAILBOX LOCKING * I have added mandatory locking of mailbox files where supported. This will cover Linux systems, in particular. * The default behavior is now to do lock-protected append on the user's system mailbox rather than using delivermail or some other MDA. (This is a performance hack.) * The autoconfigure script now looks for standard mail locations. The default mail delivery agent is used only if it can't find a mail spool directory in the standard places. FUNCTIONAL ENHANCEMENTS * When using POP3, message headers are edited so that replies won't foo up. Anything that looks like a mail ID local to the POP host gets @ and the pop servername attached to it before being appended to the user's mailbox or passed to an MDA. * I have implemented daemon mode. * I have added a lock check to ensure that there is only one popclient running per user, and a --quit option to kill the currently running one. DOCUMENTATION * All changes and feature additions have been tested in actual use and are documented on the man page. * I have turned the comments in the sample.poprc into a new manual section documenting the .poprc format. MISCELLANEOUS BUG FIXES * I fixed some de-initialization bugs in pop2.c and pop3.c that led to fd leaks (these became painfully obvious when I tested daemon mode!). * I've fixed the flaky parser error messages. They turned out to be due to a misdeclaration of yytext. These are Carl Harris's change notes from prevuious releases: 3.0b5 o "From " header fix in pop2.c and pop3.c o Surpress "..." output when --stdout option specified in pop3.c 3.0b4 o alloca fix. o various diagnostic/informational message fixes. 3.0b3 o Support for retrieving only new messages from maildrop when using POP3. o Support for retrieving only the first n lines of each message when using POP3. o APOP authentication support. o Buffered socket input. 3.0b2 This is a "new features" release. o support for .poprc file. o GNU-style long options. o fixed passwords appearing in 'ps' output o support for multiple servers on one command line 3.0b1 This is mostly a test of the autoconfigure integration. Among the functions performed by the new configure script, is the ability to detect known system types, configures the mail delivery agent (MDA) correctly. This should permanently solve the problem of using something other than an MDA for mail delivery (which continues to plague Linux slackware 1.2.9). For this beta, please check the values of MDA_PATH and MDA_ARGS carefully. They should match the values found in your sendmail.cf file on the line which begins with "Mlocal". Other changes from popclient version 2.21: o no longer uses getpass() from the C library. The internal getpassword() function allows the use of long passwords. o integrated GNU getopt() for long options. Long option names will appear in a future beta. o Several compiler warnings fixed. o Fixed problems related to missing include files in Solaris port.