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. Introduction ------------ 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 situation. 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 code. Mailers ------- 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. Summary ------- 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. Thanks ------ 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) ---------------------------------------------------------- Prod 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