Tillbaka till svenska Fidonet
English   Information   Debug  
FIDONEWS_OLD4   0/37224
FIDO_SYSOP   12852
FIDO_UTIL   0/180
FILEFIND   0/209
FILEGATE   0/212
FILM   0/18
FNEWS_PUBLISH   4436
FN_SYSOP   41706
FN_SYSOP_OLD1   71952
FTP_FIDO   0/2
FTSC_PUBLIC   1693/13613
FUNNY   0/4886
GENEALOGY.EUR   0/71
GET_INFO   105
GOLDED   0/408
HAM   0/16074
HOLYSMOKE   0/6791
HOT_SITES   0/1
HTMLEDIT   0/71
HUB203   466
HUB_100   264
HUB_400   39
HUMOR   0/29
IC   0/2851
INTERNET   0/424
INTERUSER   0/3
IP_CONNECT   719
JAMNNTPD   0/233
JAMTLAND   0/47
KATTY_KORNER   0/41
LAN   0/16
LINUX-USER   0/19
LINUXHELP   0/1155
LINUX   0/22112
LINUX_BBS   0/957
mail   18.68
mail_fore_ok   249
MENSA   0/341
MODERATOR   0/102
MONTE   0/992
MOSCOW_OKLAHOMA   0/1245
MUFFIN   0/783
MUSIC   0/321
N203_STAT   930
N203_SYSCHAT   313
NET203   321
NET204   69
NET_DEV   0/10
NORD.ADMIN   0/101
NORD.CHAT   0/2572
NORD.FIDONET   189
NORD.HARDWARE   0/28
NORD.KULTUR   0/114
NORD.PROG   0/32
NORD.SOFTWARE   0/88
NORD.TEKNIK   0/58
NORD   0/453
OCCULT_CHAT   0/93
OS2BBS   0/787
OS2DOSBBS   0/580
OS2HW   0/42
OS2INET   0/37
OS2LAN   0/134
OS2PROG   0/36
OS2REXX   0/113
OS2USER-L   207
OS2   0/4786
OSDEBATE   0/18996
PASCAL   0/490
PERL   0/457
PHP   0/45
POINTS   0/405
POLITICS   0/29554
POL_INC   0/14731
PSION   103
R20_ADMIN   1123
R20_AMATORRADIO   0/2
R20_BEST_OF_FIDONET   13
R20_CHAT   0/893
R20_DEPP   0/3
R20_DEV   399
R20_ECHO2   1379
R20_ECHOPRES   0/35
R20_ESTAT   0/719
R20_FIDONETPROG...
...RAM.MYPOINT
  0/2
R20_FIDONETPROGRAM   0/22
R20_FIDONET   0/248
R20_FILEFIND   0/24
R20_FILEFOUND   0/22
R20_HIFI   0/3
R20_INFO2   3249
R20_INTERNET   0/12940
R20_INTRESSE   0/60
R20_INTR_KOM   0/99
R20_KANDIDAT.CHAT   42
R20_KANDIDAT   28
R20_KOM_DEV   112
R20_KONTROLL   0/13300
R20_KORSET   0/18
R20_LOKALTRAFIK   0/24
R20_MODERATOR   0/1852
R20_NC   76
R20_NET200   245
R20_NETWORK.OTH...
...ERNETS
  0/13
R20_OPERATIVSYS...
...TEM.LINUX
  0/44
R20_PROGRAMVAROR   0/1
R20_REC2NEC   534
R20_SFOSM   0/341
R20_SF   0/108
R20_SPRAK.ENGLISH   0/1
R20_SQUISH   107
R20_TEST   2
R20_WORST_OF_FIDONET   12
RAR   0/9
RA_MULTI   106
RA_UTIL   0/162
REGCON.EUR   0/2056
REGCON   0/13
SCIENCE   0/1206
SF   0/239
SHAREWARE_SUPPORT   0/5146
SHAREWRE   0/14
SIMPSONS   0/169
STATS_OLD1   0/2539.065
STATS_OLD2   0/2530
STATS_OLD3   0/2395.095
STATS_OLD4   0/1692.25
SURVIVOR   0/495
SYSOPS_CORNER   0/3
SYSOP   0/84
TAGLINES   0/112
TEAMOS2   0/4530
TECH   0/2617
TEST.444   0/105
TRAPDOOR   0/19
TREK   0/755
TUB   0/290
UFO   0/40
UNIX   0/1316
USA_EURLINK   0/102
USR_MODEMS   0/1
VATICAN   0/2740
VIETNAM_VETS   0/14
VIRUS   0/378
VIRUS_INFO   0/201
VISUAL_BASIC   0/473
WHITEHOUSE   0/5187
WIN2000   0/101
WIN32   0/30
WIN95   0/4289
WIN95_OLD1   0/70272
WINDOWS   0/1517
WWB_SYSOP   0/419
WWB_TECH   0/810
ZCC-PUBLIC   0/1
ZEC   4

 
4DOS   0/134
ABORTION   0/7
ALASKA_CHAT   0/506
ALLFIX_FILE   0/1313
ALLFIX_FILE_OLD1   0/7997
ALT_DOS   0/152
AMATEUR_RADIO   0/1039
AMIGASALE   0/14
AMIGA   0/331
AMIGA_INT   0/1
AMIGA_PROG   0/20
AMIGA_SYSOP   0/26
ANIME   0/15
ARGUS   0/924
ASCII_ART   0/340
ASIAN_LINK   0/651
ASTRONOMY   0/417
AUDIO   0/92
AUTOMOBILE_RACING   0/105
BABYLON5   0/17862
BAG   135
BATPOWER   0/361
BBBS.ENGLISH   0/382
BBSLAW   0/109
BBS_ADS   0/5290
BBS_INTERNET   0/507
BIBLE   0/3563
BINKD   0/1119
BINKLEY   0/215
BLUEWAVE   0/2173
CABLE_MODEMS   0/25
CBM   0/46
CDRECORD   0/66
CDROM   0/20
CLASSIC_COMPUTER   0/378
COMICS   0/15
CONSPRCY   0/899
COOKING   33421
COOKING_OLD1   0/24719
COOKING_OLD2   0/40862
COOKING_OLD3   0/37489
COOKING_OLD4   0/35496
COOKING_OLD5   9370
C_ECHO   0/189
C_PLUSPLUS   0/31
DIRTY_DOZEN   0/201
DOORGAMES   0/2065
DOS_INTERNET   0/196
duplikat   6002
ECHOLIST   0/18295
EC_SUPPORT   0/318
ELECTRONICS   0/359
ELEKTRONIK.GER   1534
ENET.LINGUISTIC   0/13
ENET.POLITICS   0/4
ENET.SOFT   0/11701
ENET.SYSOP   33945
ENET.TALKS   0/32
ENGLISH_TUTOR   0/2000
EVOLUTION   0/1335
FDECHO   0/217
FDN_ANNOUNCE   0/7068
FIDONEWS   24159
FIDONEWS_OLD1   0/49742
FIDONEWS_OLD2   0/35949
FIDONEWS_OLD3   0/30874
Möte FTSC_PUBLIC, 13613 texter
 lista första sista föregående nästa
Text 3987, 256 rader
Skriven 2010-03-13 10:34:42 av Michiel van der Vlist (2:280/5555)
Ärende: FSP-1029
================
Hello All,


=== Cut ===
**********************************************************************
FTSC                             FIDONET TECHNICAL STANDARDS COMMITTEE
**********************************************************************
Publication:    FSP-1029
Revision:       2 (draft 3)
Title:          Serial number allocation in Mesage IDs.
Author(s):      Michiel van der Vlist

Date:           25 February 2010
----------------------------------------------------------------------
Contents:       0. Status of this document.
                1. Introduction.
                2. The problem.
                3. The solution.
                4. The standard.
                   4.1 The serial number file.
                   4.2 Name and path of the serial number file.
                   4.3 Issuing numbers using the serial number file.
                   4.4 Upon successful open and lock.
                   4.5 Upon failure to open the serial number file.
                   4.6 Upon failure to obtain a lock.
                   4.7 When the serial number file is corrupted.
                   4.8 A note on "increasing by at least one".
                5. References.
                6. Contact Info.
----------------------------------------------------------------------

0. Status of this document
--------------------------

  This document is a Fidonet Standards Proposal (FSP).

  This document proposes a Fidonet standard for the Fidonet
  community.

  This document is released to the public domain, and may be used,
  copied or modified for any purpose whatever.


1. Introduction
---------------

  FTS-0009 mandates that serial numbers used in FTN Message IDs be
  unique on a given system within a period of three years. FTS-0009
  does not mandate any particular way to achieve this goal. It does
  not even offer suggestions, It leaves it entirely to the
  implementation.

  This has led to software developers using a variety of methods,
  using the system clock is the most popular, but random number
  generators and checksums over the message header and/or body are
  also used.


2. The Problem
--------------

  The problem is that none of the above mentioned methods is really
  satisfactory as it can not guarantee that no duplicate message IDs
  will be issued. Issuing duplicate message ID's is undesirable as
  echomail processors and netmail trackers may discard or quarantine
  messages as dupes.

  Using the system clock was good enough when computers were slow and
  single tasking. With computers getting faster, only issuing a new
  serial number every second however is too coarse. Fast systems can
  create tens if not hundreds of messages a second. Using a finer
  scale, i.e. hundreds of a second or even thousands only shifts the
  problem to the future. Some day we shall see systems that can
  spew out more than one message in a millisecond.

  Multitasking systems have worsened the problem. Not only can one
  application give out the same number twice, Another application
  running in another task may give out that same number as well!

  Using random number generators and/or checksums is simply wrong.
  There is just no guarantee no duplicates will occur. It is roulette,
  Some day the same number will come up again.


3. The Solution
---------------

  The obvious solution is to use some sort of "sequencer" that
  is shared by all applications running on a given system that
  issue message ID's. The idea is simple, it is the same as
  drawing a "ticket" from the ticket dispenser in the Post Office.
  No two tickets in the machine have the same number and only
  one customer can draw a ticket at the time. If the machine is busy
  dispensing a ticket to a customer, the next customer has to wait.
  This way it is ensured that no two customers get the same number.

  There are many ways to implement the equivalent of such a system
  for FTN serial numbers, but it will only work properly if all
  applications use the same method. Hence the need for a standard.


4. The standard
---------------

  4.1 The serial number file.

        The serial number file is a plain text file who's first
        line is exactly eight characters long. It holds an eight
        digit hexadicimal number, representing a 32 bit unsigned
        integer. Leading zeros may not be omitted.
        It is recommended that applications write this number in
        lower case. When reading it must be treated as case
        insensitive.

        The eight character line may optionally be terminated by a
        line terminator. The file may contain additional lines
        containing additional information, thus making it possible
        to use this file for supplementary future standards. Appli-
        cations issuing serial numbers according to this proposed
        standard shall only consider the first eight bytes of the
        serial number file.


  4.2 Name, extension and path of the serial number file.

        The default name of the serial number file shall be MSGIDNR.NXT.
        It is recommended that the name be configurable.

        The path pointing to the file MUST be configurable. For ease
        of configuration it is strongly recommended that the
        implementation supports reading the path from the environment
        variable SERNRPATH.

  4.3 Issuing serial numbers using the serial number file.

        An application needing to issue a serial number for a message
        ID shall attempt to open the serial number file and lock the
        file. Or alternatively, if the OS permits partial locking, lock
        part of it with a minimum of the first 8 bytes.


  4.4 Upon successful open and lock.

        On a successful lock it shall read the 32 bit hexadecimal
        number from the file and use that as the next serial number
        to be issued.

        After having issued this number it shall increase the serial
        number by at least one and use that for the next serial
        number. When the programme is done issuing serial numbers,
        it shall write the last issued serial number increased by at
        least one to the serial number file, release the lock and
        close the file.

        Alternatively if more than one serial number is to be issued
        in one run, the programme may write the next to issue serial
        number to the file, unlock and close it; and open, lock and
        read it again between issuing serial numbers so that other
        applications can obtain a serial number in the meantime.


  4.5 Upon failure to open the serial number file.

        Upon failure to open the file because it does not exist, the
        application shall attempt to create it. What method is used
        to "seed" the serial number generator is up to the implemen-
        tation. The system clock might be a good suggestion, but this
        document shall not attempt to mandate any particular method.


  4.6 Upon failure to obtain a lock.

        Failure to obtain a lock normally means another application
        has locked the file. The logical course of action would be to
        wait for the lock to clear and retry. Limiting the time to a
        maximum before aborting and/or supplying a user abort option
        are at the discretion of the implementation.

        Another reason for failure to obtain a lock can be that the
        application is running on a single tasking system that does
        not support file sharing and locking. (F.e. plain DOS, SHARE
        not loaded.) If the application is aware of being run in a
        single tasking environment, either by autodetect or by
        a configuration switch, it can of course skip the locking
        and just read and update the sequence number file.

  4.7 When the serial number file is corrupted.

        When the serial number file contains characters other than
        0-9, A-Z or a-z in the first eight positions, this is an error
        condition. One possible course of action would be to re-
        initialize the serial number generator. Another would be to
        abort or prompt the sysop for intervention.
        All methods have pros and cons, so this proposal does not
        impose any specific course of action to resolve this error and
        leaves it to the implementation.


  4.8 A note on "increasing by at least one".

        Simply starting at zero and increasing the serial number by
        one for each number issued would be perfectly valid and
        satisfy FTS-0009. That is, provided  all applications running
        on a given system were to use this method, use the same serial
        number file and the serial number file is never lost.

        In practise these ideal condition may not be met so using a
        more sophisticated method of increasing the serial number than
        just increasing it by one may be preferred. Using the system
        clock to prime the number generator in case the serial number
        file is absent and to use the larger value of the system clock
        and the last number incremented by one for the next number may
        give reasonable insurance against the loss of the serial
        number file.


5. References
-------------

  [FTS-0009] A standard for unique message identifiers and reply chain
             linkage.
             Jim Nutt, 17 Dec 1991.

  [FSC-0083] A proposed standard for message IDs on FTN systems.
             Jonathan de Boyne Pollard. 17 June 1995.


6. Contact Data
---------------

  Michiel van der Vlist
  Fidonet:  2:280/5555
  E-mail:   pa0mmv at vrza dot org
  E-mail:   administrator at ftsc dot org


History
-------

   Rev.1, 20040206: Initial Release.
                    Principal author Michiel van der Vlist.

   Rev.2, 20100225  Par. 4.1: 1st paragraph changed for more clearity.
                    Made some small textual changes.
                    Added section about corrupted sequence number file
**********************************************************************

=== Cut ===




Cheers, Michiel

--- GoldED+/W32-MINGW 1.1.5-b20070503
 * Origin: http://www.vlist.org (2:280/5555)