mouse@mcgill-vision.UUCP (09/05/87)
In article <917@looking.UUCP>, brad@looking.UUCP writes: > [re domaining schemes] > There is nothing wrong with having multiple logical addressing > schemes. There should be such. > But one consistent scheme must always be present. That's the scheme > you use when you try and "figure out" how to mail somebody. > If you wanted to get mail to somebody, what would be easiest to work > out? I maintain that this is the geographical scheme. Usually if > you know one thing about somebody, it's the town they live in. Not true anymore and less true in the future. There are already a handful of people I know for whom I have no idea what city/town they live, work, or do anything else in. These are people I reach electronically. With greater electronic connectivity, I expect this phenomenon to make itself felt more and more. You, for example, I think of as brad@looking.uucp, and I have to think, or ask for the map entry for looking, to find out you are in Waterloo. Nor do I care particularly where you are geographically, unless and until I have to do something (like sending paper mail) which requires such information. > You need this if you want to mail a postal letter, Necessarily. We need some way of providing a globally unique address for each potential recipient. For paper mail, the means that has been chosen is a normal name plus a street address (when this system was developed, not much else was workable, though now we could use computers to set up a registry allowing other schemes). For electronic mail, a scheme involving hierarchical things called "domains" has been chosen. (A precursor, still hanging around, involved specifying a path, as if I had to address a paper letter to you with Brad Templeton 1234 Somewhere Street via Crossing Avenue in Waterloo via Main Artery in Waterloo via Waterloo Main Post Office via .... via Montreal Main Post Office via Main Artery in Montreal via Crossing Street in Montreal this letter being sent from 4321 rue Qpt). > or find their phone number. Again, only because that's the only way it has been set up. It could be set up differently with the technology we now have, but there's been no sufficiently compelling reason to tear down what we have. > Even if you already know the number, the town is coded within it. From that point of view, once you have a domain address the town is coded within it as well, since it necessarily refers to some machine, which necessarily is in some location. This will break down when we get multiprocessor machines with the processors in widely different places, but let's not worry about that just yet. I forsee a day when there is one world-wide computer, a colossal distributed machine, but that is some time off. > This has nothing to do with the routing underneath. This is just how > I think we usually associate people. Only because that's how we are used to doing it. It does not have to be this way, and I think that we should try some other way, if only because progress is made only when we try new things. > The one, consistent scheme that everybody understands is geography. Again, only because that's the only one our society uses universally, because until very recently it was the only one we had the technology to support. > Organizational and company related schemes are haphazard and > difficult to follow. I suspect this is because they have, so far, been constructed very ad-hocly(!). > For far down the line, a geographical domain structure is the only > way to go. Now, while the number of email users is small, we can use > other schemes as we like, but they will damn us later. I fail to see why the user-visible address should have anything to do with physical location. The numeric address (for Internet mail) or routing path (for UUCP mail) necessarily is physical-location-based because communication channels tend to tie physically close machines together. (This is beginning to break down, witness uunet's connectivity.) However, this is what nameservers (for Internet mail) or UUCP maps (for UUCP mail) are for: to map the user-visible addresses into the means to reach that address. > Think about it: If you could redisign[sic] the Canada Post > addressing scheme right now (ignoring Postal Codes), Zip codes (I refuse to stumble over "Postal" codes, just as I always call all small Band-Aid-style bandages "Band-Aid"s even though they strictly aren't) are slightly disguised domain addressing, with geographical domains (natural enough, when geographical locations are being addressed). > and things like routing and post offices were not a problem, how > would you do it? This is how email should be addressed, too. I fail to see why. As you say, there is no problem until the number of email users gets to the same order of magnitude as the number of postal addresses. So let us postulate a world in which every home has a computer; let us suppose these are networked in some way. Now let us suppose you have a flame/compliment/question (or some other sort of message) about something. Let us suppose it is the rear-view mirror on your new car. Why should you have to ask where the head office of the car maker is and send your letter there? Let the computer do the grunge work: address your letter to the car maker and let the computer look up the physical, numerical, whatever address. *This* is how *all* mail should be addressed. However, we do not have the technology (OCRs &c) to do it reliably for paper mail, and it would be a humungous conversion effort even if we did. But with email we have no conversion effort, or rather the conversion effort is no worse one way than another, and we *do* have the technology (MX records and UUCP maps). der Mouse (mouse@mcgill-vision.uucp)