To LUGNET HomepageTo LUGNET News HomepageTo LUGNET Guide Homepage
 Help on Searching
 
Post new message to lugnet.generalOpen lugnet.general in your NNTP NewsreaderTo LUGNET News Traffic PageSign In (Members)
 General / 5188
5187  |  5189
Subject: 
member id's: simple numbers or something more?
Newsgroups: 
lugnet.admin.general, lugnet.general
Date: 
Sun, 4 Jul 1999 05:01:57 GMT
Highlighted: 
(details)
Viewed: 
105 times
  
In lugnet.admin.general, "Todd Lehman" <lehman@javanet.com> writes:
In lugnet.admin.general, Ben Roller writes:
[...]  That way we could all be given any old number and the
part that we would see would be a username of some sort.  I'm sure
that systems like ebay have numbers, but heck if I know what my number
is because I just use my username.  [...]

I've yet to see a username-system that (a) handles collisions in a
reasonable way if the usernames are system-assigned, or (b) doesn't lead
quickly to garbage if the usernames are user-chosen.

Hmmm, hmmm, hmmm.  Awright, maybe it's worth taking a major 180 here and
taking a closer look at non-numeric member-id's (i.e., usernames).

Personally, I think usernames are nothing but a major source of headaches
once a user population grows above a certain point, but maybe together we
can devise something that won't be too restrictive but also won't be too
abuse-prone.  Like I said, I've never seen a username-system that handles
collisions reasonably or doesn't lead to garbage if users have full control
over their username/member-ID.  But I guess that's no reason to give up.

Anyway, the basic idea behind member-ID's is this:  For the foundations of
unique member identities, each person needs a unique ID of some sort.  That
could be a simple number, or a sequence of letters, or a combination of
letters and numbers, or whatever.  The challenge is to choose a system which
allows a certain amount of personal expression, but doesn't encourage (or
hopefully even permit) crap.  We don't want LUGNET to turn into another AOL
(for example) in terms of usernames.

The ID's need to be unique and permanent because they form the basis of,
among other things, URLs which represent a person's homepage within the
community, and tacit cognitive connections which build over time as people
get to know one another.  For example, someone who sees the letters "cjc"
somewhere today is likely to think of Mike Stanley if they exchange email
with him often.  Or someone who sees the letters "moz" somewhere today is
likely to think of Chris Moseley if they know him by that nickname.

So the unique ID is a necessity from an internal mechanism point of view,
but it's not something that can ever be totally hidden from view, because
whatever form the ID's take (numeric or alphanumeric or whatever) the ID's
must inevitably appear in URLs.

Let's take a look at both ends of the spectrum and then a closer look at a
few things in the middle of the spectrum.  Perhaps something in the middle
represents a reasonable compromise between the two extremes.

At one end of the spectrum, there is a system which assigns simple counting
numbers -- like 7, or 55, or 1234 -- on a first-come first-serve basis,
starting at 1 and counting upward.  I think this is the way ICQ works, and
it's simple and effective.

At the other end of the spectrum, there is a system which allows people to
choose whatever "screen name" or "handle" or "username" they please,
provided that no one else has yet obtained that name.  This is the way AOL
works, and it's also simple and effective.

However, both extremes have deep fundamental problems.

At the purely numeric extreme (i.e., ICQ), there is no hint of who someone
is just by their number.  And even if you know someone by their number,
numbers are typically relatively easy to forget -- especially when trying to
keep a few dozen or hundred of them straight.

At the ad-hoc alpha-numeric extreme (i.e., AOL), there is sometimes a decent
hint at who someone is by their username/handle, but really only if they
have chosen to use their real name, or if they have used a reasonable
nickname that makes any sense.  AOL-style names are surely more difficult to
forget than numbers, but they are hideously prone to garbaging-up.

Here are some examples of problems with the AOL approach.  I don't know what
the character-length-limit is for AOL screen names, but I think it's more
than eight.  However, even eight characters is plenty to represent anything
from the childishly cutesy...

   acidburn, phyrefox, defcon5, neozero, immortal, genepool, pubert, etc.

to the disgorgingly trite...

   k00l1, in10siv, l8ralg8r, 2cool4u, sk8rd00d, me2me2me, win95sux, faqewe,
   bumsnifr, ob1ken, ds9rulez, etc.

to the potentially sexually, racially, or religiously offensive...

   hot4u, 6of9, mastrb8r, 13incher, gotohell, satan666, nigrhatr, killfags,
   etc.

(The above names are purely fictitious.  Any resemblance to actual screen
names or handles is a coincidence.)

Now, if people can specify whatever name they please, how can this type of
garbage be prevented in a completely automated way?  I'm not sure that
it can.  People are still infinitely more creative than machines.  That
means there has to be at least -some- set of restrictions, whether imposed
via machine (automatic) or via human intervention (manual), in order to pass
or fail each username request.  Human intervention is not really a serious
time issue, but it may potentially border on some form of censorship.

One partial solution might be to disallow numbers altogether, or to disallow
the use of numbers as letter-substitutes ("l8r", "k00l", "in10se", etc.),
but still to allow things like "whg3" (that's one way to write William H.
Gates III -- eek! :-).  But I don't think that's a full solution.

In the interests of diplomacy, certain character sequences such as "lego",
for example, must be prevented, unless someone's actual real-life name
actually contains those letters.  (According to www.switchboard.com, there
are more than 300 people in the U.S. with the last name Lego.  There's even
someone in California and someone in Oregon with the last name Lugnet. :-)

So far so good?  No...

What if someone other than Tim Courtney or Brandon Grifford attempted to
acquire the username of "zacktron"?  How would something like that be
detected and dealt with?  Or what about someone other than Tom McDonald
attempting to acquire the username of "radiotitan"?  Or what about someone
other than Simon Denscombe attempting to acquire the username of "carbon60"?

Conclusion:  It just isn't mechanically feasable to allow anywhich arbitrary
username to be chosen.

So what's the next best thing?  Say, maybe something that involves someone's
real-life name but still allows a modicum of flexibility and creativity?

One thing might be to allow any sequence of letters taken in order from
someone's full real-life name.  For example, (let's use LarryP, because I
don't think he'll mind), Larry could go for any of:

   lp, lar, larry, larryp, lpien, lpieniazek, etc.

or he could simply use his initials (although I don't know what his middle
initial is, so I can't put that one in the list).  Let's see, what would I
do?  I'd probably go for something like one of:

   tsl, todd, toddl, tlehman, toddlehman, etc.

and perhaps Suzanne would go for something like one of:

   suz, srich, suzanne, etc.

So that actually seems like a modus vivendi to me.  Sandra Linkletter of RTL
could even keep her cool handle "slink".

(What would YOU pick for your name?)

Now, are there any loopholes or problems in this scheme?

One potential problem is, ironically, three-letter nicknames (for example,
"lar" as shown above).  I believe there is actually a Larry A. Rosler in the
computing community -- so not only might his familiar nickname also be
"lar", but his initials are "lar".  Once a username is chosen or assigned,
it is permanent for lifetime, so LarryP wouldn't ever be able to sell his
member-ID of "lar" to someone such as LarryR.  That's one potential problem.

Another potential problem is due to the permanence of the member ID's:  if
someone marries or divorces and changes their last name, and if they used
their old last name in their member-ID and wanted to change it to use their
new last name, then they're S.O.L.

A third potential problem is much more complex and socio-psychological.
Under the scheme described above, Larry Pieniazek could have "lar" and Paul
Gyugyi could have "gyug", but Chris Moseley could not have "moz" and Joseph
Gonzalez could not have "gonzo" (because of s=>z and a=>o).  This would be
unfortunate because all four of those nicknames are something those four
people have used publicly, and all four of them are sense-making nicknames.

OK, so if the rules were bent slightly so that Chris Moseley could have
"moz" and Joseph Gonzalez could have "gonzo", then is that bending enough?
What other near-border cases are there?

Of course it makes sense that Roberts could always be Robs or Bobs;
Jennifers could always be Jennys or Jens; Michaels could be Mikes; Anthonys
could be Tonys or Ants; Fredriks could be Freds or Riks; and Williams could
be Wills or Bills, etc.

Ahem -- say again on that last one?  What about someone named William
Williamson?  Could he go by "billbill"?  (I saw a "billbill" on the net
once.)  If so, could John Williams go by "johnbill" or would he have to use
"johnwill" or some other thing?

More food for thought:

Should Asrun Kristmundsdottir be permitted the member-ID of "kristmas"?
Should Nick Holdbrooks be permitted the member-ID of "ickna" (I assume
that's Piglatin for "Nick")?  Should Richard Mussler-Wright (more famously
known as "Weird Richard") be permitted the member-ID of "weirdrichard"?
Should Ed Jones (famously known as Ed Boxer) be permitted the member-ID of
"edboxer"?  Should David K.Z. Harris be permitted the member-ID of "zonker"?
The list of borderline cases goes on and on...

(I have opinions on all of the above, but I'll keep them to myself for the
time being...)

Some interesting three-letter-initial combintations, BTW:  Alan B. Clegg
could be "abc"; Allan R. Martin could be "arm"; and Sean O'Brien could be
"sob".  Those are all kind of interesting.  :)

Anyway, I think something like this (above) -- a scheme which is still
highly objective, and fair, and only a very tiny bit subjective -- would
give a large amount of flexibility, and a lot of understandability and
memorability, and yet still avoid the garbage types of names that can crop
up in an unregulated system -- all without doing anything that could be
construed as censorship.  It's true that some people would definitely not
get the member-ID that they would most favorably desire, but in any system
-- even an unregulated one such as AOL's -- there's never such a guarantee
because someone could have taken your favorite name before you.

Well, that's probably enough brain-dumping for now...

[followups to lugnet.admin.general]

--Todd

p.s.  To anyone who reads this post and feels I've really gone out to the
looney farm this time, well, all I can say is that I actually find this
topic endlessly fascinating!  :-)



Message has 5 Replies:
  Re: member id's: simple numbers or something more?
 
(...) I totally agree. (...) Hey watch the AOL bashing :') But since you brought up AOL usernames - AOL allows the cability of 5 different screen names for the same account. This allows familys to have a name for each memebr of a family. Not to (...) (25 years ago, 4-Jul-99, to lugnet.admin.general, lugnet.general)
  Re: member id's: simple numbers or something more?
 
(...) Not sure where I stand on all this or if it really matters to me, but I thought I would mention this. All (or most, anyway) of my current login names that I actually have to type on a regular basis are cjc. Yeah, I like the initials because (...) (25 years ago, 4-Jul-99, to lugnet.admin.general, lugnet.general)
  Re: member id's: simple numbers or something more?
 
Univeristies do the username made from name thing all the time. Maybe you could look at some of those systems. Of course they all do it differently, but I'm so dang used to pschemp, I would want to use that. It beats the schempppat I got at another (...) (25 years ago, 4-Jul-99, to lugnet.admin.general, lugnet.general)
  Re: member id's: simple numbers or something more?
 
Need to read more but seems to me that any mechanistic scheme needs an escape clause. if we go with letter IDs, then have the automation be relatively harsh and inflexible, but allow appeal. I know, I know, you don't want to be the decider, because (...) (25 years ago, 4-Jul-99, to lugnet.admin.general, lugnet.general)
  Re: member id's: simple numbers or something more?
 
Did I really need to quote? I'm sure most everyone knows what's going on. Anyway, here's a possible sollution that I'm keen on (please hear the newbie out....) I sign up for a membership and, since the system advances member numbers (...) (25 years ago, 4-Jul-99, to lugnet.admin.general, lugnet.general)

Message is in Reply To:
  Re: Allocation of member #'s
 
(...) Because of URLs that contain them. (...) I've yet to see a username-system that (a) handles collisions in a reasonable way if the usernames are system-assigned, or (b) doesn't lead quickly to garbage if the usernames are user-chosen. --Todd (25 years ago, 3-Jul-99, to lugnet.admin.general)

112 Messages in This Thread:
(Inline display suppressed due to large size. Click Dots below to view.)
Entire Thread on One Page:
Nested:  All | Brief | Compact | Dots
Linear:  All | Brief | Compact

This Message and its Replies on One Page:
Nested:  All | Brief | Compact | Dots
Linear:  All | Brief | Compact
    

Custom Search

©2005 LUGNET. All rights reserved. - hosted by steinbruch.info GbR