rayan@utegc.UUCP (10/19/87)
[ Warning: This message is 10k, all of it prose. Grab some warm chocolate... ]
As part of the domain registration application documents (what a
mouthful!), there will be a list of contact points for submitting the
applications. The way this is intended to work, is that there will be a
central location that acts as a clearing house (the CA domain
authority) and each network will provide a liaison function for its
members. The CA domain authority will be CDNNET headquarters for now,
although in the long run this will probably move to a government body
of some sort (e.g. NRC). Having such a central clearing house is
necessary for concurrency control, to act as the authority in case of
problems, and to coordinate dissemination of routing data. The point of
encouraging interaction through a network-specific mechanism is to
decrease the load on the CA authority, and allows each network to cater
to its needs and procedures. For example, in the UUCP world it would be
advantageous to coordinate between the domain registry and the UUCP map
data. In other networks, they have their own routing data formats and
procedures for maintaining the data.
The purpose of this message is to start a bit of discussion (in private
mail if people wish) about how we should set up this liaison function
in the Canadian UUCP community. In the very short term, what I need to
come up with is an E-mail and (Paper) post address as the UUCP contact
point for submitting domain registration applications. Beyond that,
nothing is decided until there appears a consensus of what to do.
By way of background, I should mention that the functionality
envisioned is something similar to the UUCP Zone in the U.S., which
interfaces with the DDN NIC (Defense Data Network, Network Information
Center) for all UUCP sites wishing to register a domain under the
.COM/.EDU/.GOV/... toplevel domains, which cannot do it themselves by
virtue of being on the ARPA/NSF Internet. In our case, CDNNET HQ is
the analogue of the DDN NIC. There is a provision for unaffiliated
(with a network) organizations of registering directly with CDNNET HQ
for a yearly fee (planned to be 50$ or so). CDNNET will liaison (with itself :-)
for its members, at no additional cost to them, and NetNorth will do the
same I believe.
I should also clarify the situation regarding personal/private sites.
The CA authority does not want to register individuals' home PC at this
time. The intent for such sites is that they either gather together in
associations (or clubs, etc.) which can register a domain, or hang off
of registered domains using existing addressing kluges. The only way to
integrate such sites would be through kluges in the domain hierarchy,
which is deemed undesirable. The problem of assigning electronic
addresses to individuals (independent of location) is something that
will become more acute in the next few years, but it is not one we can
solve. (It is a global problem and needs to be addressed (hmm,
unintentional pun there...) by globally adhered-to standards.)
Now, there are several matters to be settled:
1. How can we do it?
There are several possibilities I can think of... let me list them:
a- we don't do it at all, and everyone just goes straight to CDNNET HQ.
b- we use the existing mechanism of the U.S. UUCP Zone.
c- we use an existing Canadian Unix organization (/usr/group/cdn).
d- we set up our own formal organization (incorporate'n all) to do it.
e- we get the Canadian UUCP Map coordinator(s) to do it.
f- we set up a good basis from which we can wing it.
I will comment on all of them (so you can see why I'd prefer f), but first let
me mention some criteria I'd like to see fulfilled:
- continuity of service (this relates to fees, to be discussed below)
- a certain level of technical competence
- a certain level of accountibility to the UUCP community in terms
of activity and use of fees.
So, my commentary:
a- I'm not at all sure they'll like that, in fact probably the opposite.
It'd also severely complicate the little matter of getting good routing
data for the .CA domains into the maps. Not to mention reflect poorly
on our self-esteem.
b- This is a possibility, since they (in the person of Mark Horton) have
expressed a willingness to help/coordinate in such matters, subject to
negotiations. There are disadvantages in that they don't have much of an
incentive, and that we wouldn't have much (any) control over how things
are run. Besides this there may be administrative difficulties of
various kinds (e.g. connectivity and interaction with CA authority,
currency problems, and it being a U.S. organization). There has been
a certain amount of griping at the handwaving Mark did when asked to
explain what the UUCP Zone fees were going to.
c- Not to slight any such organizations, but if (and where) they exist
they seem to have a different focus than to provide this kind of
service (at least from what I've seen). It would probably be premature
to rely on them for our purposes.
d- This is probably what should happen eventually in some form, to look
after the interests of the Canadian UUCP community vis-a-vis all the
things that are happening that will affect us. However, I don't think
we are ready for such structure as yet. This is a possible evolution
for /usr/group/cdn and/or whatever we come up with now.
e- I'm it, and its a thankless enough volunteer job as it is. This would
be ideal from the point of view of coordinating with the UUCP map.
Personally, I think I've put a lot of time in over the last 3 years
to the community benefit, and I'm not happy about adding to the load
merely "out of the goodness of my heart".
f- "winging it" is perhaps a misnomer in this connection; what I mean is
there isn't a real formal organization to actually do the work. This
message is slightly delayed because I wanted to investigate what kind
of local infrastructure support could be made available to our project.
It appears the local powers are willing to lend a hand in this, as I
shall describe below.
The following elements are necessary to run a liaison function as intended:
- someone to do the work.
- a computer account to do it in/from.
- a mail drop (paper mail) for letters containing fees.
- administrative support to process fees and manage the monies in an account.
I asked the director of the Computer Systems Research Institute here
about this matter, and to which degree they could help. They are (and
very gracious of them, I might add) willing to set up an account,
provide a mail drop, and set up and manage money account(s) and process
fees as long as the volume remains reasonable ("a couple of hundred
checks a year or so"). The computer account, labour, and any
administrative costs would be paid for out of the fees. They are aware
this is a community service item, and therefore administrative costs to
us would be 0 as long as the load is reasonable. As well, the computer account
would be charged at some factor (as yet unknown, but low) above that used
for primary research users around here. It is understood that for now I
would provide most of the labour, but there will be continuity should I
leave or keel over dead (which probably means I'll be training local staff
once things become routine and/or I get tired of it).
This takes care of the operational side of things, the remaining item is:
2. Fees.
Why have fees at all? Several reasons:
- It costs money to store, maintain, and disseminate the necessary data.
If not directly, then certainly indirectly in hidden or opportunity costs.
- To ensure continuity, it is necessary to ensure that running or sponsoring
such a service is not a losing proposition, neither directly in financial
terms, nor in (my fav. phrase) opportunity cost.
- To ensure proper procedures are followed in the applying organizations.
- As a nuisance fee to keep the crackpots away.
I don't think it is realistic to NOT have a fee for these reasons.
Assuming we do have a fee, what should it be? Well, it should at least
cover expenses, i.e. computer account, labour, and any administrative expenses.
Estimating these figures is very hard to do due to lack of experience.
I'll take a rough stab at it here:
computer account:
say disk usage of 30k per subdomain. This covers the variable usage for:
original application form, munged ditto, corresponding map entries for the
subdomain (different from normal UUCP link info), and storing correspondence
with the subdomain and with the CA registrar about the subdomain.
In addition, this includes a share of the storage needed for maintenance
programs and various other data (pathalias databases, log files, etc.),
which depends on how many registrations are handled.
cpu time? I don't really know... the UUCP map maintenance account takes
a couple of cpu hours, on average, per month on a VAX780. I'd guess
somewhere between 5 and 10 cpu minutes per subdomain while it is being
registered, and a few seconds per month after that.
Assuming we'll be charged at a factor 2 over minimum rates, the computer
account charges will be approx:
2 * (0.2 * 22$/hr cpu + 12*30*2 * 0.0085 $/blk-mon) = 2 * 10.52 = 21$/year.
For comparison, applying commercial rates on the CSRI machines would make
it a factor of 12 instead of 2.
labour:
Assuming an hourly rate of 25$, and somewhere between 10 and 30 minutes spent
in total per subdomain (the actual time depends on a lot of things, primarily
the complexity of the site), that gives approx. 25/3 = 8$. This could be
paid as some factor (1.1 or so) times the session time on the computer account,
at some hourly rate. The factor is intended to take into account any
non-electronic activities.
Administrative costs and other expenses cannot be estimated. Their total
will probably be non-zero, but cost/subdomain should be negligible.
So, in total, with estimates that may seem a bit on the high side, the total
cost per year is on the order of 30$ per subdomain. Therefore I suggest the
fee be set at 30$/year <= fee <= 50$/year.
Soo, to conclude this very long message, my proposal is:
We make use of CSRI facilities to provide the necessary computer and
administrative support (and postal drops, etc.). For now, I'll coordinate
things with the UUCP community, CSRI, and the CA authority, until the
need or desire arises to pass on the operation to someone else trained in
doing it. The yearly fee shall be X$ (what do you think X should be?),
and will be managed by the local burearcracy (e.g. checks would have to
be made payable to UofT, but end up in an account set up for the purpose).
The fee money will be used to pay expenses as outlined above.
If it turns out after a year or so that the fee is too high or low, we can
readjust it then (to avoid problems, I think it is better to start out with
a high fee).
Comments invited.
rayanbrad@looking.UUCP (10/19/87)
When the original domain schemes were first proposed on the arpanet tcp mailing lists, I (and others) thought that they were simple and elegant. A way to get mail easy to use without endless administrative hassle. They have turned out to be just the opposite. Constant bickering. "Official" organizations. Long proposals. Silly fee structures that have to do with the administering of the mail rather than the moving of it. Counter proposals. Suggestion of handing it to the NRC, for christ's sake. What a shame. It could have been nice. Perhaps the only solution at this time is for somebody else to break away totally and set their own new standard. It's worked in the past. -- Brad Templeton, Looking Glass Software Ltd. - Waterloo, Ontario 519/884-7473
lamy@utegc.UUCP (10/19/87)
| A way to get mail easy to use without endless administrative hassle. I fail to see how administrative hassles for the administrators relates to the ease of use for the users. The only way domain-based mail can prosper and indeed become easy to use is if it is done right. Indeed, getting mail set up properly is currently a major undertaking (I have seen 5 sites come on the net recently, with various levels of interconnectivity), and my impression is that 30$ is a small fee if it results in *any* improvement in the current mail addressing picture. 30$ is roughly the price of someone fighting with mail addressing for an hour... Let me take you through my current .signature, for fun lamy@ai.toronto.edu this is what it should have been, or so I hoped when I was naive. lamy%ai.toronto.edu@relay.cs.net this it has to be when coming from the ARPA side, since some sites still think that relay.cs.net is a strange host name with dots in it. lamy@ai.toronto.cdn ai.toronto.edu works with EAN all right, but mail from Montreal to Toronto then goes through Vancouver (UBC) and Boston (relay.cs.net) {uunet,watmath}!ai.toronto.edu!lamy ah, this is getting reasonable. uunet and watmath understand domains lamy@utorgpu.bitnet (not for real) this does not exist. but since (most?) bitnet hosts have never heard of domains, the fact that technically ai.toronto.edu is part of the officially registered toronto.edu is not of any help. I can send mail to all bitnet sites, but can't receive any from a lot of them. So currently, I would need to have an account on a "real" BitNet host to be able to get BitNet mail... The only way this is going to get better is if the organizations get together and keep talking. This process seems to be underway: - There are talks of BitNet and CSNet merging their network centers. - There are rumours of a TCP/IP BitNet backbone. - There is an experimental Canadian Internet (TCP/IP over EAN's X.25) - More Canadian sites are going on the Arpanet (McGill and the Montreal institutions on the CRIM network are; Toronto has X25Net capabilities and has received ARPA approval for a slot in an IMP; UBC and Waterloo sound like plausible candidates as well) - X.400 is catching in Europe (this is the protocol used by EAN) - All ARPA and BitNet mailing lists will likely be UseNet newsgroups in the near future. Can you imagine the fun with return addresses unless some order is imposed? And so, in a day and age where finally e-mail and news systems seem to be converging towards greater interconnection, we are left with this puzzling suggestion: | Perhaps the only solution at this | time is for somebody else to break away totally and set their own new | standard. I rest my case. Jean-Francois Lamy lamy@ai.toronto.edu, lamy@ai.toronto.cdn AI Group, Dept of Computer Science lamy%ai.toronto.edu@relay.cs.net University of Toronto, Canada M5S 1A4 {uunet,watmath}!ai.toronto.edu!lamy
rayan@ai.toronto.edu.UUCP (11/22/87)
This is basically to summarize the comments I got... The bottom line is that the operation will be run as proposed, and fees will start out at 50$/annum. Before the summary of comments, I should just mention that indeed the charging factor applied to the account will be as I assumed, so no numbers will change (good thing too, because the next higher charging factor would have more than doubled the numbers -- thanks go to CSRI for their cooperation). Anyway, apart from Dave Martindale's philosophical gripe posted here, there were no other negative comments, and half a dozen or so positive ones. Thanks to Henry Spencer, David Trueman, Ian Darwin, Dave Sherman, Eric Carroll, and Michael Herman, for their responses. A couple of the comments are worth broadcasting here: I think it's a good proposal. $50 sounds like a reasonable fee. It wasn't clear to me to what extent the "expenses" include paying you at $25/hour -- was that the intention? I have no problem with that either. Provided the service works and you report on the finances to the net once every 6 months or whatever, that should be fine. In particular, you will be taking the risk of making less if the work proves to require more than you estimated. But do make it clear, in public, that the net would be hiring you for this job. If we hear screams, complaints, or offers from others to do it for less, let's have that out in the open too. For the record, I support your proposal. Yes, that was the intention. staff rates are a bit over 20$/hour, and my opportunity cost for work inside UofT corresponds to the rate I assumed (and I can only work a small N hours per year). This isn't exactly a money- making opportunity. The report on finances suggestion is fine with me; I was thinking about something similar already. I would suggest that you push for a formal organization to be created to insulate uoft and csri, as I suspect that the work load and cost will quickly become unmanagable. UofT and CSRI are insulated in a sense. What exists is a sort of gentlemans agreement for CSRI to provide some support. It is understood that if the load becomes too high, they will receive appropriate reimbursement. I thought about a separate entity to run this thing, but I don't think it is justifiable yet, from the point of view of workload at least. I have been slightly concerned about the legal problems that might crop up, but as long as I'm not the CA authority, and behave responsibly, one would hope the world won't place silly obstacles in the way. And I would approve of a connection cost (zone fee) of ~$100 cdn a year. After all, I happily pay $150 US for my .COM registration, and would likely pay more if I had to. A higher price will make the pc users think twice. and One comment, however. Although it is probably impossible to administer, it would be nice if larger companies/sites paid a larger fee. I think the US project has some sort of hierarchy to address this. Yes, it would be nice. I'm not sure if it is worthwhile though. The real costs are not really related to the size of the organization, but rather to the complexity of its interface to the world. Certainly there is a correlation. One could of course make a higher fee voluntary (...) but the accounting people want to see something in writing that says how much must be paid. Are there any other thoughts on this matter? Maybe we could have two versions of the instructions, and send the one with the higher price on request ;-) In the absence of a workable and fair scheme, I think $50 across-the-board will do for now, and we can think up something new for next year, if needed. The documents will be distributed widely tomorrow (Monday 23rd Nov). This is almost exactly a year since the discussion started in earnest. rayan
rayan@ai.toronto.edu.UUCP (11/23/87)
In article <1987Nov22.145125.10426@jarvis.csri.toronto.edu> I write:
# Anyway, apart from Dave Martindale's philosophical gripe posted here, there
# were no other negative comments, and half a dozen or so positive ones.
Oops, I remembered the originator wrongly, it was Brad Templeton, not
Dave Martindale. My apologies to both of them. The "philosophical gripe" wasn't
anything for anyone to want to disavow though -- from what I recall of it,
to a certain extent I agree and sympathize. There are other considerations
however.
Sorry.
rayan