Document: FSC-0039
Version:  04
Date:     29-Sep-90

                      A Type-2 Packet Extension Proposal
                       Mark A. Howard 1:260/340@FidoNet

  Status of this document:
  This FSC suggests a proposed protocol for the FidoNet(r) community,
  and requests discussion and suggestions for improvements.  Distribution
  of this document is unlimited.

  Fido and FidoNet are registered marks of Tom Jennings and Fido Software.
  FTS-0001 is a copyrighted work of Randy Bush.

  This document serves two major purposes.  The first is an attempt to define
  and document the Type-2 packet which is widely in use in FidoNet today.
  Although FTS-0001 defines the structure of a Type-2 packet, the natural
  evolution of our network, mostly with regards to addressing methodology,
  has made it necessary to utilize hitherto unused portions of the header
  to insert Zone and Point information.  Also, it has become apparent that
  some of the existing fields are not large enough to accomplish their
  original tasks.

  The second is to propose a simple mechanism to allow FidoNet to begin to
  utilize advanced mail packing techniques.  It is quite apparent that while
  Type-2 has served us faithfully for some time now, the evolution of our
  network in terms of technical and physical complexity has caused us to
  consider more efficient and functional ways to pack mail.

  It should be made clear that with the exception of the Capability Word,
  Capability Word Validation Copy, ProductCode(hi), and Revision(minor),
  which are proposed extensions to the Type-2 packet header, this FSC is
  an attempt to correctly document existing practices with regards to the
  insertion of zone and point info by at least three mail processors in
  use today.

  The Type-2 Header (Where's the Zone?)
  Although FTS-0001 has recently been updated to reflect the use of some of
  the areas in the packed message header for zone data, at least two other
  methods for inserting the zone information have been adopted, making it
  necessary to provide support for both formats in all of the zone aware mail
  processors.  The end result is more code, and redundant information in the
  packet header.

  This has presented a problem in logistics, as it is difficult to consider
  the project of updating mail processors using one type to use the other.
  As sufficient indentification is provided, in the form of the product code,
  to determine the expected location of the zone information, and because
  code already exists in most of the mail processors to overcome this, this
  proposal does not attempt to suggest that one method be used over the
  other, rather the intent is to attempt to document the extensions in use,
  and the products involved.

  See the accompanying chart and cross-reference.

  The Product Code
  Based upon the current rate of requests for product codes from the FTSC,
  it is probable that the Product Code byte will not be large enough to
  accomodate all of the codes required.  While it is not reasonable to
  expect that all Type-2 processors will eventually check the hi-order byte
  proposed here, it is likely that 'current' mail processors will.  This
  can be nothing but benefical, as it will force users to upgrade their
  mail processors to a product which will as a minimum, support Type-2
  with Zone and Point extensions, and quite possibly, processors that will
  utilize more advanced mail packing techniques, making Type-2 extinct once
  and for all.

  The Capability Word  (How do we GET there from here?)
  Everybody would like to see more efficient and functional ways to pack and
  exchange mail.  Several Type-3 message bundle proposals exist, but none
  really address a problem which must be solved first.  The problem is that
  since FidoNet is a hobbyist network, no demands can be placed on any one
  sysop to upgrade or change their bundling software.  Because of this, it
  is necessary to consider strategies which allow for the existence of Type-2
  bundlers in the network topology.

  Considerable advantages can be realized, however, between systems that
  consent to use advanced bundling techniques.  One way to do this is to
  simply send netmail to all of your connecting systems, saying "Hey, I've
  got a Type-3 bundler now, how about you?"  This could become quite
  tiresome, and does not represent much of an improvement on the current

  What would be desirable is a network that would 'upgrade itself'.  Given a
  situation where mail processors of various capabilities will exist in a
  network topology, the goal is to provide a mechanism whereby two links can
  determine and utilize the most efficient bundling method to use, in a
  manner transparent to the sysop.

  For instance, let's say that the FTSC releases the Type-7 All New Singing
  and Dancing bundle format.  Well, your current version of SlingToss can
  only support Types 2, 3, and 5.  One of your downlinks gets a new version
  of MailMangle which can support Types 2, 3, 4, and 7.  Well, it is quite
  obvious that since you and he are exchanging 4 megs of mail each night,
  and it's an overseas call, that it would be in your interest to obtain a
  new version of SlingToss which can support Type 7.

  Note that this is *optional*.  Because both processors can support Type-3,
  they will continue to exchange Type-3 mail quite happily, even though
  MailMangle is happily advertising the availability of Type-7.  Even your
  downlinks which are still using stone-age Type-2 processors will be fine,
  as SlingToss will always export Type-2 bundles when no higher capability
  can be determined.

  So, after dashing off the check to the author, your new version of
  SlingToss comes in the mail!  You rush over to your system, and install it.
  The next time SlingToss exports mail to the MailMangle system, it says
  "Hey!  I can now support Type 2, 3, 5, and 7!  So, whattya got?"  This is
  no skin off MailMangle's nose, he's had Type-7 for quite a while, and he
  begins to export Type-7 bundles to SlingToss.  "It's about time.", he says.

  Now, this scenario is made possible by implementing a 'Capability Word' in
  the present and future packet headers.  The Capability Update mechanism
  depends on several assumptions:

  1)   Any Advanced Capability Bundler *MUST* be capable of receiving and
     faithfully processing Type-2 bundles.  Hopefully, the inbound packets
     will be in the new format proposed by this document, but then again,
     this is not an exact science.  What this means is that it is likely
     that some packets may arrive with the Capability Word (CW) set to 0.
     In this case, Type-2 is assumed, assuring compatibility.  The only
     caveat is that it is conceivable that some obscure mail processor
     uses the location proposed for the CW for other arcane purposes.  This
|    can detected through the CWValidation Copy, which is byte-swapped and
|    compared with the CW at that time.  If a mismatch is found, a CW of
|    type 0 is assumed, and a Type-2 bundling method is used.

  2) An Advanced Capability Bundler, hereafter referred to as a Type-N
     Bundler, must have a method to store and maintain the node-by-node
     capability information.  This can be done many ways, and in fact
     several processors already have begun to maintain node information
     outside of that found in AREAS.BBS, mostly to implement pre-arranged
     alternate compression methods.  In a text configuration file, you
     might see the following:

     ;       Address      Comp    Send  LastCW ; Comments
     Node    1:260/340    ZIP     Auto  7      ; Auto detect & upgrade
     Node    1:135/20     LZH     3     2,3,7  ; Always send Type-3
     Node    1:           ARC     2     0      ; Stone-Age processor
     Node    1:135/4      ---     Auto  7      ; Sent me netmail
     Node    1:           ---     0     0      ; Don't send CW

     In this example, the fields are:

     Address - downlink address.  Note that this is not necessarily
               relative to echomail, only, it is possible to append
               information to the node database as netmail packets are
               receieved from different addresses.

     Comp    - desired mail compression method.

     Send    - Auto - automatically determined maximum common packing
                      method to use.  Automatically update to LastCW
                      when packing.

     LastCW  - Last CW received from remote system.

  3) A Type-N Bundle will always advertise it's capabilities in the CW
     regardless of the type being sent.  As shown in the above example,
     it allows Type-N processors to automatically track the capability
     of your system.  Again, in cases where a stone-age processor is
     being used, this field will be ignored, and in the unusual event
     that it is not ignored, and is somehow harmful to the far system,
     the Type-N processor can be configured to send a CW of 0.

  The format of the Capability Word is designed to support up to 15 future
  bundle types, and is bit-mapped to facilitate the easy determination of
  the maximum common level supported between two nodes:

                 msb           Capability Word               lsb
  Node Supports  ------------FTSC Type Supported----------------

                  U 16 15 14 13 12 11 10  9  8  7  6  5  4  3  2

  2, 3, and 7     0  0  0  0  0  0  0  0  0  0  1  0  0  0  1  1
  2, 3, and 5     0  0  0  0  0  0  0  0  0  0  0  0  1  0  1  1
  2 (this FSC)    0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  1
  Stone Age**     0  0  0  0  0  0  0  0  0  0  0  0  0  0  0  0
                  +--- Indicates UseNet RFC-822 capability

                  ** - a mismatch in the CWValidation Copy also
                       produces a CW=0.

  In this example, the Type-N bundler would first compare the remote CW
| and the byte-swapped remote CWValidation Copy, and check for a mismatch.
  Prior to the compare, the MSB of the CW's are masked, as this bit is
  reserved to indicate whether the mail processor is capable of also
  accepting UseNet RFC-822 bundles.  Following the MSB mask, and bit
  comparison, if they do not match, a remote CW of 0 is assumed.  If they
  match, the Type-N processor would AND the local and remote CW words,
  obtaining a CW expressing the Types which are in common to both systems.
  The most significant Type will be used, by default, but note that this
  assumes that new bundling Types will be increasingly more efficient or
  in some way more beneficial.

  Because this may not always be the case, there should be a method provided,
  as illustrated above, to override the automatic upgrade should this become
  the case.

  The MSB of the CW is used to indicate whether the mail processor can accept
  UseNet RFC-822 bundles or not.  It is a separate indicator, and not intended
  to be used as a part of the above comparison, however can be used to also
  advertise RFC-822 capability to other systems.  Since RFC-822 is 'set in
  stone', there is no need to assign more than one capability bit.

  It might seem somewhat limiting to only consider the possibility of 15
  different future bundling methods, but it is my opinion that the careful
  use of a 'Sub-Type' byte in the packet header can allow for the variations
  on a single theme, and that proposals for new formats should be evaluated
  by the FTSC to determine whether sufficient benefit can be realized in
  the implementation of the given format, prior to assigning a new type

  It is quite clear to me that should this concept take hold, that the
  logical place to store node capability data is in the local nodelist
  database, or an index-associated data file.  As above, it is necessary
  to generate Type-2 packets for whatever purpose, unless it is known
  by prior association, that the far mailer can accept other types of
  packets.  It is also quite reasonable to assume that a nodelist flag
  could be assigned to advertise the CW for a given node, which the
  native mailer nodelist compiler could then immediately determine the
  preferred bundling method for any given node in FidoNet.

  Another possibility would be to pass a capability advertisement in the
  extensible portion of a handshake protocol, which may or may not already
  exist in FidoNet.

  The approach suggested previously in this document suggests the use of
  a text configuration file, but it is quite obvious that many benefits
  can be realized through the use of the nodelist, including the use of
  additional flags to indicate the preferred compression method, etc.

  This document has been created in an attempt to define a method to allow
  the future expansion and enhancement of FidoNet technology mail processors
  and mailers, and in a way that is the least disruptive to existing mail
  operations.  The intent is to provide for an environment that is as open,
  and extensible as possible.

  The mechanism described should allow many different types of processors
  (FTSC-registered) to exist in the network at once, and to provide an
  environment which is designed to operate at it's maximum efficiency
  wherever possible or practical.

  Revision 2 of this document was produced to implement suggestions made
  primarily by Jan Vroonhof, who suggested the use of the CW Validation
  Copy.  Jan presented this idea in his FSC-0048, along with other concepts
  relating to the correct indentification and handling of zone and point
  addressing.   This document sanctions the improvements to the CW as
  recommended, but does not address or support the other extensions
  recommended in FSC-0048.

  To Ward Christensen, creator of XModem and BYE.

     Tom Jennings, who started this whole mess.

     Joaquim Homrighausen, for lots of good ideas, and motivation.  Here's
                           another Lamborghini to work on.

     Wynn Wagner, Oliver McDonald, Roeland Meyer, Andrew Farmer, Claude
     Warren, Jan Vroonhof, Bob Hartman, and Vince Perriello, who all
     contributed in some way to the creation of this document, mostly
     through their messages in NET_DEV.

  Type-2 Packet Format (proposed, but currently in use)
    Field    Ofs Siz Type  Description                Expected value(s)
    -------  --- --- ----  -------------------------- -----------------
    OrgNode  0x0   2 Word  Origination node address   0-65535
    DstNode    2   2 Word  Destination node address   1-65535
    Year       4   2  Int  Year packet generated      19??-2???
    Month      6   2  Int  Month  "        "          0-11 (0=Jan)
    Day        8   2  Int  Day    "        "          1-31
    Hour       A   2  Int  Hour   "        "          0-23
    Min        C   2  Int  Minute "        "          0-59
    Sec        E   2  Int  Second "        "          0-59
    Baud      10   2  Int  Baud Rate (not in use)     ????
    PktVer    12   2  Int  Packet Version             Always 2
    OrgNet    14   2 Word  Origination net address    1-65535
    DstNet    16   2 Word  Destination net address    1-65535
    PrdCodL   18   1 Byte  FTSC Product Code     (lo) 1-255
  * PVMajor   19   1 Byte  FTSC Product Rev   (major) 1-255
    Password  1A   8 Char  Packet password            A-Z,0-9
  * QOrgZone  22   2  Int  Orig Zone (ZMailQ,QMail)   1-65535
  * QDstZone  24   2  Int  Dest Zone (ZMailQ,QMail)   1-65535
    Filler    26   2 Byte  Spare Change               ?
| * CapValid  28   2 Word  CW Byte-Swapped Valid Copy BitField
  * PrdCodH   2A   1 Byte  FTSC Product Code     (hi) 1-255
  * PVMinor   2B   1 Byte  FTSC Product Rev   (minor) 1-255
  * CapWord   2C   2 Word  Capability Word            BitField
  * OrigZone  2E   2  Int  Origination Zone           1-65535
  * DestZone  30   2  Int  Destination Zone           1-65535
  * OrigPoint 32   2  Int  Origination Point          1-65535
  * DestPoint 34   2  Int  Destination Point          1-65535
  * ProdData  36   4 Long  Product-specific data      Whatever
    PktTerm   3A   2 Word  Packet terminator          0000

  * - extensions to FTS-0001

  Ofs, Siz are in hex, other values are decimal.

  Zone/Point Aware Mail Processors (probably a partial list)
    Code Name - Uses QOrg/QDstZone Orig/DestZone Orig/DestPoint
    ---- ----------- ------------- ------------- --------------
    0x0C  FrontDoor  Reads/Updates      Yes           Yes
    0x1A  DBridge        ?????          Yes           Yes
    0x45  XRS        Reads/Updates      Yes           Yes
    0x29  QMail           Yes          ?????      Not point-aware
    0x35  ZMailQ          Yes          ?????      Not point-aware
    0x3F  TosScan    Reads/Updates      Yes           Yes