[138] in I/T Delivery
Re: 2/27/98 Delivery Team Leaders Meeting Minutes
daemon@ATHENA.MIT.EDU (Theodore Y. Ts'o)
Mon Mar 2 20:28:07 1998
Date: Mon, 2 Mar 1998 20:28:01 -0500
From: "Theodore Y. Ts'o" <tytso@MIT.EDU>
To: wdc@MIT.EDU
Cc: delivery@MIT.EDU
In-Reply-To: Bill Cattey On the Road's message of Fri, 27 Feb 98 12:15:12 EST,
<9802271741.AA02713@MIT.MIT.EDU>
Date: Fri, 27 Feb 98 12:15:12 EST
From: wdc@MIT.EDU (Bill Cattey On the Road)
Ted T'so: Suzana visited Integration team to look at other search
engines than Harvest. Requirements review is key. A real discovery
effort is required.
Susan pointed out that I confused two separate issues in my round-table
report. The CWIS team approached us for two different topics. The
first was the search engine, and there we gave more tactical advice,
such as looking at different engines (not just the one which the
resource development office was pushing), etc.
The second topic was the "counter" service, where the CWIS team would
like us to offer a web counter service which would be reliable enough so
that Technology Review could use it to charge advertisers for money when
users both (a) viewed their pages, and (b) when measuring clickthroughs.
Although technically, it's not hard to run a (usually reliable) web
counter, it brings up all sorts of business issues.
For example Technology Review is sellings ads to cost-recover, and I/S
would have to put in resources just so that TR can rake in money. Is
this appropriate? What are the service level agreements that are in
place here? If the service (which we would be providing at no cost to
TR) breaks, so that TR can't charge their advertisers money, who bears
the business liability? In the web advertising business, the
advertisers often demand the ability to audit the hit logs --- what are
the privacy implications here?
It was this latter issue which we recommended to should require a more
formal discovery process. (To be sure, there are some customers who
aren't planning on using counters for commercial functions, and simply
for curiosity reasons or to justify the resources their department was
putting into the web pages. Although here too, reliability issues in
terms of user expectation also come into play here. If you want a
mostly available, "best efforts" counter, you can use the one which SIPB
provides, or ask Jeff or me about the personal web counters which we run
on each of our respective private workstations. But the CWIS team
indicated that wasn't good enough; they wanted something that was more
reliable. How reliable? Is it worth paging someone at 3am to fix it?
etc.)
In any case, my apologies for the confusing these two issues during my
round-table report.
- Ted
