From Leonard Erickson · 1996-04-10 20:17 · #73166 in FREE25
In reply to NET_DEV by David Appleby
-=> Quoting David Appleby to Leonard Erickson <=- LE> I disagree. If the new format has extra info, then translating to the LE> *old* format will be easy, because you just leave out the extra info. LE> (for instance type 2+ and 2.2 PKT files have info that a type 2 system LE> neither needs nor wants). DA> And how do you convert the message back? There's not much point in DA> only being able to convert one way. If systems that can handle the DA> new type packets are connecting further along the chain they should be DA> able to recreate the new packet from the information in the Type-2 DA> packet. When you start out, you'll only be using the "new" format with the few systems you connect to that support it. So "converting back" is a pretty silly idea for the majority that is still on the old packets. When it gets well established, then only the leaf nodes are going to be using it, and again, there's no point in "converting back". So the only time being able to "convert back" from the old format to the new format would make any sense would be during the period when there are roughly equal amounts of old and new systems. That's going to be a *very* short window. Consider, if the new format doesn't catch on, then it will never get there. If it does, then it will take very little time to go from "most don't use it" to "most do use it". DA> For example, if the new packet type had the address of the system that DA> created the message in the header, in addition to the address of the DA> system originating the current copy of the message, but did not DA> require an origin line there would have to be some way of putting that DA> information into the Type-2 packet that is created in backwards DA> translation so that it could be recreated when the translation goes DA> forward again. Please note that packet formats and echomail requirements are *seoerate* items. So leaving out the origin line means that you've nbot only redefined packet formats, but echomail standards as well. Even so, you've answered your own question. Convert from old to new, you'd take the Origin line, and put the address in the appropriate header field. Converting from new to old, you'd check and if there wasn't an origin line, you'd add one, using info from the appropriate header field. There are *only* two classes of info here. Info that exists in both formats (regardless of whether it is stored the same way) and info that only exists in the newer format. I just showed how to handle the former. And for the other, you throw away the extra info. It's *only* of use if you wind up doing the reverse translation somewhere. And for a number of very good reasons, it's best to arrange your network so that you never have mesages getting translated A->B->A, regardless of whether A is new format or old format. The Internet learned that a long time ago with ASCII/EBCDIC conversion. LE> We can handle other changes in similar ways. The mailers don't care LE> *what* kind of files they transfer. You just need something to process LE> them. If nobody writes a handler for your system, you fly a flag saying LE> that you can't. And folks who send stuff to you have to honor it. DA> Which would be another flag for the software authors to ignore. Not DA> to mention the fact that there is no agreement between zones on what DA> nodelist flags are allowed and disallowed. Software authors don't always ignore flags, and it's possible for sysops to manually set their configs too. As for interzone agreement on flags, it'd help if we had an IC. DA> don't know about you, but Squish over here will quite happily send an DA> archived message to a node flying the MN flag if I have a default LE> I'm sure that you can *override* the default on a system by system LE> basis, can't you? If not, then all it takes is an MN system complaining LE> and *you* would have to change software. DA> I can override the default but if the default is set to send ARCMail DA> to all nodes unless their is a specific override in the configuration DA> file the system will send ARCMail bundles to a system that flying the DA> MN flag. The defaults are changed by discussion between me and my DA> links, not by various flags in the nodelist. So it can be configured for. The fact that it isn't more automatic is just due to nobody making an issue of it. LE> BTW, the fact that the version 7 nodelist omits info is a problem for LE> the software that uses it, not for the rest of the network. DA> Does your software recognize the MN flag and refuse to send ARCMail DA> bundles to those sites? I'm not a hub. DA> Does your software recognize the Gxxx flags DA> when routing mail to another networks gate? I'm not sure. I don't *need* to worry about GUUCP, since I've got an Internet address. And the only other G flag I've ever noticed isn't in the nodelist anymore. DA> Does your software handle DA> the ISDN flags when they appear alone in the nodelist without any DA> other modem capability flags and realize that it can't connect there? If I care to tell it to. FD is *very* flexible on modem flags. Up to and including the ability to say "don't call that". --- Blue Wave/DOS v2.30 * Origin: Shadowshack (1:105/51)