The World of Computer Software

home *** CD-ROM | disk | FTP | other *** search

/ The World of Computer Software / The World of Computer Software.iso / 130b_31.zip / STATUS.TXT < prev

Wrap

Text File | 1993-01-05 | 28KB | 713 lines

QM beta test. Limited distribution test version. Testers can share this with associates but I don't want to see or hear about it being up for file request anywhere. Folks you pass it to are "your helpers". They funnel reports thru you. As always, users of this software are testers. There are risks to echomail and other files on your system. Please be careful while you have fun... 1.30/b31 01/05/93 - Incoming bundles using -lh4- compression are processed the same as -lh5- (via Lha) instead of failing. - Zero length and some unpacking errors were creating trouble. This one might operate a little more reliably. But it might occasionally mistreat a non-mail file that's named just right. - Archive unpacks are now strictly by date/time rather than date/time within day of week extensions. 1.30/b30 10/28/92 - When compressed_mail_max_bytes wasn't used every pack operation created a new bundle. - If Dupe_History_File is set to NUL or NULL dupe checking and all dupe database operations are disabled. 1.30/b29 09/30/92 - Toss scan operations didn't look pretty when max messages was hit. The max messages and finishing packet notifications were displayed once for every message. Cut it back to showing the messages only once. - A max message condition during toss scan exited with error- level 4 (messages scanned) rather than 5 (max messages). Level 5 implies 4 so should have been used. - Add_To_Seen addresses have always been added to echo area scan counts. Since they don't actually result in mail being sent they throw off statistics and accounting programs. So this time around we'll see what folks think about making the log file Sent= value show only what's actually scanned out. 1.30/b28 09/29/92 - Added configuration parameter Compressed_Mail_Max_Bytes to define the maximum file size for archived mail. A new file will be started whenever any compressed mail bundle larger than that size is found during a pack operation. - The new archive extensions must be turned on explicitly with the New_Archive_Ext Qm.Cfg directive. 1.30/b27 09/03/92 - When a toss scan operation (vs toss scan pack) reaches max msgs scanned it no longer drops into toss only mode when finishing up the current packet. It was doing that to make sure outbound packet sizes honored max_msgs but it makes a mess in a real passthru environment. - Outbound archived mail is generated with A-Z in addition to the old 0-9 day of week extensions (.TU8 .TU9 .TUA .TUB ... TUZ) for a maximum of 36 uniquely named archived bundles a day. - .?lO files larger than 4096 bytes could be trashed. 4096 bytes were read in but the original file size was written back out when adding a new archived file name. - Unpack is a new command line operation a few folks have asked for. "qm unpack" unarchives / deletes mail archives and exits without tossing. Errorlevel 7 is returned if anything was unpacked. If unpack is used with any other verb (toss scan pack link) it will be ignored. 1.30/b26 08/06/92 - the dupe catcher is now one file rather than lots of little files. there is no conversion. go for broke. before you run this one set the qm.cfg dupes parameter to the highest number of messages you think you'll want to track. that's the number that'll be used to initialize the file. after the file is created you can lower the dupes value so new records will use it but you cannot increase it higher than the initial setting without deleting and recreating the file. this is because records in the file are fixed length and cannot be expanded. support for the dupe database requires about 4k additional memory at run time. each message tracked requires 6 bytes. tracking 2000 messages (the default) per area requires about 12000 bytes per record. that's around 8mb for the whole backbone load. decreasing or increasing the dupes parameter at initialization time will change the fixed record size overhead. The maximum capacity of the dupe database is 2048 areas. you can delete all those little dupe files and your passthru directories after you switch to this version. - the dupe database filename is defaults to qmdupes.dat in the current directory. use the qm.cfg parameter dupe_history_file if you need to override the default. - qmdupes.exe is included. it's a dupe database maintenance utility. - hide_area was causing the area: line to be fatally hidden. rather than inserting ^a the a of area was overlaid with ^z. - the year was being placed in outbound packets as an offset from 1900 rather than the full year as indicated in fts-0001. - type 2.2 inbound packets (fsc-0045 proposal) are recognized and logged correctly. the point and domain fields are not used. this is a defensive move since some software forces type 2.2 on us. - increased the standard toss buffer and max message size by 1k. the buffer capacity is now 32k and the max message size is 31k. qmlite sizing remains at 17k/16k. 1.30/b25 ?/??/92 - only the first 6 characters of packet passwords were being checked on inbound packets. outbound packets were being built correctly with 8 character passwords. this one should build outbound packets correctly. 1.30/b24 6/22/92 - after looking more carefully at the beta 23 solution for processing trashed seen-by and path lines i found yet another problem and some unnecessary overhead in the trash detector. my solution was badly broken and created more troubles than it fixed. i'm back with a bang, eh? <sigh> 1.30/b23 6/18/92 - this one does a bit better job of detecting and jumping around trashed seen-by and path lines. whenever an unexpected character is found the program skips the rest of that seen-by or path line. the result might be dupes or strange looking paths but i think it's better to bypass the trash than to trash the whole message. after all, the header was good and we found an origin by the time we got here. - added hide_area and hide_seenby control file parameters. if used these cause ^aarea: or ^aseen-by: lines to be written rather than lines without the ^a. be careful. lots of software still dislikes ^aseen-by:. this is just a little something so qm can start to "conform" to the changing faces of technical standards somewhere down the road. don't ask. 1.30/b22 - let's forget this one and see if a gentle b23 will get us rolling again. 1.30/b21 2/ 4/92 - changed a little code trying to get ready for 4d support. the change is internal in address compare routines so the transition from 3d to 4d will be less traumatic when it finally happens. - outbound attach messages didn't use the normalized aka zone in the .msg header. this should make the messages look better to seadog and fd. - changed attach message create log entry from verbose mode to debug mode. - finally got rid of all the funny screen output and converted to 'stdout' so odd screen sizes and redirection will look better. a couple of things do look different because we lost direct cursor positioning. and a couple of old old bugs in .?lo handling... - only linefeeds were being put between lines in .?lo files. this one uses cr/lf pairs. - if some other program fiddled with a .?lo and terminated it with ^z qm left the ^z in the file and appended new data after it, rendering that data invisible to the ^z dependent utility and possible even some mailers. this one changes ^z to space. 1.30/gamma 1/30/92 - duplicate of beta 20 i hope 1.30/b20 1/25/92 - trying to fix up link code. this one seems to link in more cases than b19. 1.30/b19 1/23/92 - finally did some more work on the command line link operation. This module uses reasonably safe but admittedly not terribly fast system services rather than the sometimes-dangerous but impressively fast low-level ones some of us saw in v1.02/a. Only non-passthru areas are linked. examples might include: qm toss scan pack link \ re-link non-passthru areas qm toss link / tossed during this execution qm link re-link all non-passthru areas qm link -fechotoss.log re-link areas in echotoss.log please be careful. this one fiddles with all .msg headers. if it's broken so will be your message areas. - updated clean_outbound to skip all action if no arcmail messages are found in the configured netmail area. also threw in a couple additional feedback messages and added -d (debug) command line support to display progress messages to stdout (for redirection) and to skip truncate/delete during debug operations. - yet another stab at ignoring files with archive extensions that aren't really archives. the program will now track up to 64 of these during any one execution. any others will be renamed to bad_arc.###. the best way to test this one is to copy a text file [or several] to ?.fr0 or ?.mo1 or whatever and process mail as normal. if the program makes note of them and then neither hangs, loops, nor renames them the test is successful. hangs and loops should be reported as bugs. - fixed [again] the situation where echotoss.log was not written unless qm.log was turned on. 1.30/b18 1/7/92 - seems there was still a hole in bad bundle processing. the case where a bundle was grunged after the first .pkt was not covered so the bundle and qm went into another forever loop. this time qm will rename a bundle as soon as it receives a bad status from the unpacker rather than waiting to see if any packets were extracted. 1.30/b17 1/2/92 - re-installed bad_bndl rename code. added log file entries for rename action. - added optional "noscan" to the qm.cfg netmail directive. this directs qm to never scan netmail unless explicitly requested via the qm_route.cfg netmail_scan directive. netmail v:\netmail netmail noscan v:\netmail - rewrote the seenby/path exception check to catch more (all?) illegal characters. - during a toss operation -k on the command line tells qm to delete routed echomail that's not explicitly allowed via the qm.cfg RouteThru directive. - file processing error messages now include free memory. 1.30/b16 12/02/91 - echomail scans weren't pretty in b15. found/fixed a typo. let's try again. 1.30/b15 11/26/91 - fixed the seenby/path address processor so it doesn't hang when it finds invaid characters in the net/node lists. - compressed bundle names are now generated using the primary address from qm.cfg rather than the aka in use for packets. - compressed bundle names beginning with '-' are not unpacked. - and for peter stern we're logging attach message writes but only temporarily. 1.30/b14 11/13/91 - this one has a new address parser that tries not to get upset about points (.), domains (@), and comments (;). please test carefully. - secure mode toss from a directory (as in bad_msgs) checks only the net/node regardless of check_zones in qm.cfg. - secure mode toss from a packet is changed so only the net/node is checked in packets with no zone information regardless of check_zones in qm.cfg. - first stab at leaving arcmail style outbound attach messages alone. if a netmail message has the same characteristics that qm uses to create attach messages the netmail will be skipped. - qm no longer renames an inbound file to bad_arc.### if the packing method cannot be determined. it will also no longer incorrectly rename non-mail files that happen to have a name that looks like a bundle. if the file happens to be in a recognized archive format qm will end up renaming it if no packets are extracted. - added kill_compressed_mail directive to qm.cfg. this one lets us decide whether to delete archived mail bundles before or after tossing. deleting bundles before tossing the .pkts helps with disk space and prevents having the bundle unpacked/tossed again if errors occur during the toss. the default is after for compatibility in /b14 but i'd like to release 1.30 with before. feedback? kill_compressed_mail before tossing packets kill_compressed_mail after tossing packets - packets with a filename starting with dash (-) should not be tossed. - removed debug display of shell execute and parameters. 1.30/b13 10/10/91 - we're back to the expanded capacity version with 2048 areas, 31k toss buffer (for 30k max message size), and a choice of a 32k or 64k scan buffer as our primary qm.exe. - i've included 'qmlite.exe' with 512 areas, 17k toss buffer (for 16k max message size), and a 20k (small) scan buffer. it's pretty much a match of the 1.00 configuration for folks with limited resources. - buffersize small is converted to buffersize medium so i don't get us in the same mess as last time i fiddled with buffer sizes. so the two choices are medium (32k) and large (64k). - fixed a long-standing problem with outbound zone extension processing. getting technical, a subroutine returned a local variable address without declaring the variable as static. the result was unreliable values and any number of strange errors. 1.30/b12 10/03/91 - there's something called qmb12big.exe along with this beta. it's exactly the same as the b12 qm.exe except that the toss buffer size is 31k and the max message size is 30k. if it works reliably we'll keep it. if not you have an improved version to fall back to. - changed configuration to allow up to 2048 areas again - buffersize small is not allowed this time around. it will be changed to medium if found. - can't create this or that messages were not terribly informative. tried to add a brief reason for the most common error reasons to the information logged. - the default swap numeric value is now 519 (7 + 512) rather than 7 to change the swap to disk operation to test for a network and not pre-allocate the swap file if on a net drive. 1.30/b11 10/01/91 - backed away from some of the new stuff to see if the program runs any better on a couple of systems. - message max size is 16k - max areas is 1024 - just for grins, stopped running ms exepack on the executable. - recommend leaving buffersize to default value (medium). 1.30/b10 09/26/91 - the b09 and b10 cuts complete my goal of merging the 1.00 and 1.02 capabilities into a single program. it should offer the many faithful v1.00 users a long overdue opportunity to upgrade and the [illegal] 1.02 alpha users incentive to move to a supported version. actually, this version is arriving at least 3 months early. i'd planned to add the 1.02 features (swapping, attach outbound, large scan buffer, etc) in v1.50, 2 more updates down the road. rather than think it's already been assumed... inclusion of features in a test version does not guarantee its release. i don't mind 'leaks' of the feature list but do hope such leaks will represent friendly sharing of what is currently being tested, not what is committed for release. i don't want to cause testers any embarassment by failing to release specific features you have commented on. and of course i'm not anxious to have our bug lists show up in usa today. - this is the new stuff that might please a few folks if it makes it through the final cut. - swapping for pack/unpack/dos - messages up to 30k - up to 2048 areas - optional message attach style outbound - scan buffer size 20k, 32k, or 64k. - routed echomail - outbound aka address matching - swapping as indicated in the b09 description ems/xms/file swapping can be enabled via the qm.cfg 'swap' directive. the swap routine is a public domain package. if you want to talk technical about it request exec31.zip from 1:13/13. exec.h explains the numeric values that can be passed to the routine to control it and all the return codes it uses. the optional number following 'swap' in qm.cfg is passed directly to the routine as the spawn argument. if swap doesn't work on your box we need to trade info about what's unique about your configuration. particularly dos version and ems/xms handler. and whether 'swap 4' -- the force to disk option -- works. binkleyterm uses essentially the same swap code. - 30k messages messages up to 30k (header + body + control) are tossed, scanned, and packed now. v1.00 had a 16k limit. it might be possible to raise this to 60k with a moderate effort. i'm still looking at that -- cautiously. - 2048 areas i picked this number out of the air. it should be safe for a couple of months anyway. - attach style outbound yes, this is 'frontdoor' support. if you simply add msg between send or route and the flavor in qm_route.cfg the feature will be switched on for that destination. obviously <?> this feature needs a mailer like frontdoor. binkleyterm doesn't do this. send msg hold 1:163/115 route msg direct 1:260/1 260/all 2:all also available are the 'use_flags' and 'clean_outbound' directives in qm.cfg. - buffersize the scan buffer size can be indicated in qm.cfg. this eats away the memory available for swapping if you don't use the new swap support. buffersize large <64k> buffersize medium <32k> buffersize small <20k> default - routed echomail the 'routethru' qm.cfg directive defines a list of addresses that can route echomail transparently through your system. echomail from listed addresses addressed to an address other than your own primary or aka are not checked in any way. instead they are transferred directly to the outbound. - outbound aka address matching the 'normalize_zone' qm.cfg directive triggers aka matching or 'address assumption' for outbound mail packets. in this mode of operation qm identifies itself using an aka address when it builds outbound packets. in addition, a few old things have been tweaked... - when swapping is turned off qm now bypasses the new swap routine entirely. - the swap routine now always spawns using '%comspec% /c ...' rather than first trying a direct execute. this 'fixes' the problem of arce.com failing. - the swap routine was trying to carry forward i/o redirection. that's turned off on an effort to limit problems. - qm.cfg parsing thought the %s parameter in unpack_xxx lines was a trailing comment and truncated the argument string. 1.30/b09 09/24/91 - this version has a good deal of 'new code'. test carefully with exactly the same configuration files you used with b07/b08. the most critical change is from the medium memory model to large. this should get rid of configuration related problems but does cause qm to use more memory. try turning "swap" on if you see odd pack/unpack/dos behavior. i repeat ... T E S T C A R E F U L L Y ... your detailed feedback in beta_130 is quite important this time around. - please limit your testing to your b07/b08 configurations and concentrate on locating new problems with old features. if we all do well with this one we'll try some new tricks in b10. 1.30/b08 09/20/91 - areas.bbs lookup tests areanames for a match after a lookup by crc. this should [as in areafix] eliminate areaname colissions without suffering a great loss of performance. - swapping is implemented for mail unpacking, mail packing, and qm_route.cfg dos operations via the "swap" directive in qm.cfg. if "swap" is found in qm.cfg qm will swap all but about 2k to ems or xms or disk as available. this might change but for now... swap select from ems/xms/disk in that order swap 0 never swap (same as commenting out swap) swap 1 allow ems swapping swap 2 allow xms swapping swap 4 allow disk swapping swap 7 allow ems/xms/disk swapping (the default) 1.30/b07 09/12/91 - changed areas.bbs file offset calculation so area files longer than 64k bytes can now be used without problems. - qm will [try to] create missing outbound.??? directories rather than aborting and leaving a mess when otherzone traffic is packed. - toss operation log changes. grunged bundle offset is displayed in both decimal and hex. packer product code is displayed in hex. 1.30/b06 09/05/91 - changed seenby/path parser to strip zone, point, and domain info and to try to avoid passing along duplicate seen-by entries. - added debug logging of filename info when a message create fails. - updated high water mark message to quit saying QMail and to insert NOECHO on the first line of the message body. 1.30/b05 08/13/91 - the only change is to expand the number of areas from 512 to 600. this hange stresses the program considerably so it needs to be isolated. 1.30/b04 08/09/91 - add lha and ARJ unpacking via optional qm.cfg directives: lha_unpack 0 lha e /m %s *.pkt arj_unpack 0 arj x -y %s *.pkt - the "dos" routing command is executed every time the pack process is invoked rather than only the final time. this requires that programs invoked not change message areas in a way that might affect the high message number for an in-progress operation. - duplicate checking logic ignores incoming eids. qm calculates the message crc at toss time in order to provide consistent checks and predictable results on any one system. this may occasionally allow messages grunged by errant software to pass but in turn qm no longer dupes out messages with sequential or unusual crc or date values due to differences in calculation methods. the crc is now calculated using only message header fields (from, to, subject) in order to fully remove dependence on message body. Here's the way the new crc calculation stacks up against the old... < v1.30 > < v1.00 - the old way> eid ignored used if found, otherwise... from 35 bytes or to @ first 2 words to 35 bytes first 2 words subject up to 71 bytes first 25 bytes origin not used system name & address used date/time last character not used all fields are converted to upper case for comparisons. leading "re: " strings are skipped on subject. the @ check is because of gating software that adds data. you'll notice that the last character of the 20 byte message creation date/time field is now included in the crc. that was done to try to rid us of the false dupe situation that occurs when a series of messages have identical header information except for the units digit of the seconds. Seems even when qm started paying attention to the seconds it wasn't good enough. there are 86400 seconds in a day and only 65535 possible values for the time of day stored in a 4 hex digit eid style value. so similar messages with consecutive digits in the creation seconds position had about a 50-50 chance of getting past without some external help. my try_this solution uses the seconds unit position to fix that trouble. the entire message crc routine has been rewritten for a net savings of a badly needed couple hundred bytes. watch out for new problems. - netmail pack now does minimal header validation before processing. If any of the header text fields (from, to, subj, date) is not null-terminated the message is skipped. - fixed logging routine so errors echoed to the screen are not overwritten. - yet another date fix. make sure there are 2 spaces between year and time whenever rebuilding message header date. - fixed maxmsgs errorlevel exit to exit with the documented 5 rather than 1. - changed meaningless error message when no addresses are found in areas.bbs. - fixed missed outbound scanning log data when exiting with maxmsgs. 1.30/b03 08/01/91 - shotgun update to write the origin to local outbound echomail. 1.30/b02 07/30/91 - Dropped a good deal of the new code from 1.30/b so we can get rolling. This should be a drop-in replacement for the version 1.00 or 1.00a qm.exe. - eids will not be added to passthru or local echomail traffic and will not be used if found. the dupe detection crc value is calculated for each message. - no longer zeros the seconds field of passthru messages when checking for duplicates. - no longer converts seadog/fido format dates in message header to opus format during toss. - converts unknown format dates in message header to opus format during toss if the format is not fts-0001 spec. - adds a space (for a total of 2) between the year and month in opus format message header dates if needed. - locally generated dates now have 2 spaces between the year and time. - echo toss log no longer needs to have status log configured. - changed netmail via line and tear line to show the level. ------------------------------------------------------------------ ------------------------------------------------------------------ For swapping, the swap argument must contain a combination of the following flags: 0 - turn off swapping 1 - allow EMS swap 2 - allow XMS swap 4 - allow File swap 7 - allow EMS, XMS, and File swap (default if no value is given The order of trying the different swap methods can be controlled with one of the flags 7 - EMS, XMS, File (default) 16 - XMS, EMS, File If swapping is to File, the attribute of the swap file can be set to "hidden", so users are not irritated by strange files appearing out of nowhere by adding 64 to the value. The behaviour on Network drives can be changed by adding 256 to never preallocate or 512 to not preallocate if on net. This checking for Network is mainly to compensate for a strange slowdown on Novell networks when preallocating a file. You can either add 256 to avoid allocation in any case, or 512 to let the prep_swap routine decide whether to do preallocation or not depending on the file being on a network drive (this will only work with DOS 3.1 or later).