Association of Research Libraries (ARL®)

Several messages pick up various topical threads that arose earlier in the discussion.

Date: Sun, 21 Aug 94 10:59:24 -0600 From: Paul Ginsparg 505-667-7353

Date: Sat, 13 Aug 94 18:25:38 EDT From: "Stevan Harnad" harnad@princeton.edu

The generality and adaptiveness of the www superset is impressive! But ftp/gopher also has a PROVIDER-side argument: In text-only, non-tech (non-Tex) disciplines the probability of a successful subversion knocking down the paper house of cards is MUCH higher if authors need merely store their ascii texts rather than convert them or learn html (trivial as it is). -- S.H.

i wasn't clear enough, and this is an important point: of course, OF COURSE, www can be used to transmit plain text (this is a trivial corollary of my statement that it is a superset of gopher). after all, i'm using it to transmit .tex, .dvi, .ps, etc. -- it can transmit anything, bytes are bytes. more specifically, if an http server sees a file with e.g. a .txt (or other unrecognized extension), it tells the client that plain text is on the way and the client presents it unformatted (i'm surprised you haven't encountered this before). that is why gopher is dying out worldwide (indeed it is only naive confusion and misinformation on the above issues responsible for keeping it afloat even this long). everything gopher does, www does just as well or better (including automatic indexing of pre-existing directories). anyway, just a matter of time -- makes little difference to worry about it on way or another.

Once the subversion has had its effect, we can convert them to the virtues of hypertext, etc. (But your point on the generality of www is taken!).

and now the point of the hypertext project becomes clear -- we do transmit all this non-html via www, but these have all been network dead-ends. so rather than wait forever for some group of ncsa undergrads or whomever to reproduce a satisfactory typesetting environment within these primitive html browsers, we've taken the shortcut of adding html capabilities to our preferred medium and its browsers. (in particular that means i've been able to reprocess all pre-existing tex source in the new mode, and internal linkages are produced automatically, with no modification of the underlying .tex )

Paul Ginsparg

Date: Tue, 23 Aug 1994 08:27:14 EDT Subject: Re: ftp vs. gopher vs. www From: Rich Wiggins wiggins@msu.edu To: Multiple recipients of list VPIEJ-L VPIEJ-L@VTVM1.BitNet

that is why gopher is dying out worldwide (indeed it is only naive confusion and misinformation on the above issues responsible for keeping it afloat even this long). everything gopher does, www does just as well or better (including automatic indexing of pre-existing directories).

This claim is not quite true. The Web does not embrace the Gopher+ extensions, which have never been popular among HTTP/HTML aficionados, and are not implemented in Mosaic and its descendants.

Gopher+ provides a mechanism for alternate typing of documents. The theory is that information providers might offer documents in a variety of ways and intelligent clients might help users select among them. Web folks feel that multiple document types are handled just fine "their way" and that alternate views can be coded as part of the HTML.

But Gopher+ also provides a mechanism for named attributes of documents -- the sort of stuff like the date of the last update, author's e-mail address, etc. This is the sort of "meta-information" that is talked about interminably in IETF and Web discussion groups. Gopher+ included a mechanism for adding such attributes as of early 1993. Even in the Gopher community, though, it seems it isn't widely exploited. There are conventions for some meta-information in HTML, and no doubt discussions will lead to real standards.

The "yes there is Gopher+ but it is useless" discussion has been carried out elsewhere, and probably wouldn't be helpful here. Most new announcements of online services seem to be coming from the Web side. In general, I view Gopher as part of a progression from FTP to hierarchical menus with nice titles to Web-style hypermedia. Mosaic paved the way for the Web; now we need is bandwidth to deliver all those inline logos.

Rich Wiggins, CWIS Coordinator, Michigan State University

Date: Tue, 23 Aug 94 13:45:43 -0600 From: Paul Ginsparg 505-667-7353 ginsparg@qfwfq.lanl.gov

I agree the gopher/www quibbling is trivial

as was pointed out in message you just forwarded -- that whole discussion has been carried out through a multi-hundred message thread on comp.infosystems.gopher and comp.infosystems.www (probably still continues).

(although again misses that "in-line" logos are not necessary to www servers, they are a choice -- and i was careful to make them purely elective for everything i did, which included checking that everything worked fine from a vt100 using lynx, so that my less well-off colleagues are not left behind).

but it all remains irrelevant to the issue of costs of journals that we try to focus on -- whatever the final delivery protocol (and it may in five years be something other than what we have now, though most likely some generalization that encompasses it). but as you frequently point out, i'm here "preaching to the converted."

Paul Ginsparg

Date: Mon, 22 Aug 94 08:45:39 EDT From: Janet Fisher FISHER@MITVMA.BitNet Subject: Odlyzko on Net Capacity and Citation Frequency

It has been interesting to read Bernard Naylor's responses to the "Esoteric Publication" discussion, as well as the comments on his comments. They triggered some questions in my mind which I would like to pose.

If authors were to be charged for access to the eyes of the audience rather than readers for subscriptions, where would this money come from? The university's research budget? The library's materials budget? The government? If the university, how much publication would the available money allow for the university's faculty members? Would it be enough? It would be interesting to take some sample universities of various sizes, look at the relevant budget areas that might be applied to publication of "esoteric" research and determine if that amount would cover the amount of publication being done by that faculty. How much per page would that allow?

Regarding the lack of information about number of times an article is cited, I wonder if there is information from Paul Ginsparg about how many times articles in HEPnet are looked at, and how many times they are downloaded? What are the typical numbers? What is the pattern of usage? What percentage occurs in the first year of publication, and what percentage occurs in the second and third years after "publication"? What do we know about long-term usage?

I believe that the number of times cited is not entirely a measure of the article's usefulness or interest to the community. It seems like often researchers would review an article but not cite it in their own research because it is not exactly on point to the current argument. That doesn't mean the article is of no use to that researcher. Maybe I'm being naive...

I think it would be helpful to get as much concrete data as possible. Especially about where the money for these author charges is likely to come from if subscription charges are done away with for "esoteric" publications. Is that money there? If not, will authors pay out of their own pockets?

Janet Fisher MIT Press

Date: Wed, 24 Aug 94 19:25:47 EDT From: "Stevan Harnad" harnad@princeton.edu

Here is another long-distance reply from Paul Ginsparg at Les Houches. Henceforth I will refer to his archive as the "Los Alamos Physics E-Print Archive," because, as he indicates, "HEP" (high energy physics) is now far too narrow a descriptor for the scale it has reached in physics. Below, Paul replies to Janet Fisher's questions about usage levels. One VERY important point to note is that these statistics are for what readership levels in esoteric research WOULD be if they were not constrained by the admission price imposed by paper publication and its associated costs and practices.

Stevan Harnad

Date: Wed, 24 Aug 94 16:37:25 -0600 From: Paul Ginsparg 505-667-7353 ginsparg@qfwfq.lanl.gov Subject: Re: Author-Side Electronic Page-Cost Subsidy for Esoteric Publication

Date: Mon, 22 Aug 94 08:45:39 EDT From: Janet Fisher < FISHER@MITVMA.BitNet> Subject: Odlyzko on Net Capacity and Citation Frequency

Regarding the lack of information about number of times an article is cited, I wonder if there is information from Paul Ginsparg about how many times articles in HEPnet are looked at, and how many times they are downloaded?

what is the difference between "look at" and "download"? (in any politically correct client/server environment, one must download in order to peruse -- note that one does not do remote logins, that's why these servers have no problem handling large amounts of traffic)

What are the typical numbers? What is the pattern of usage? What percentage occurs in the first year of publication, and what percentage occurs in the second and third years after "publication"? What do we know about long-term usage?

indeed i have statistics going back to aug '91 when the original hep-th started (by the way HEPnet is a misnomer [not my terminology], since the e-print archives cover far more than just high energy physics; and moreover HEPnet is the name of the [completely unrelated] high energy physics DECnet that started in the early 80's.)

these statistics are subtle to assess for a variety of reasons. for example, there are places that simply download everything either for local printing or local caching; so there's an automatic lower bound on the number of times these things are accessed, which is not necessarily a measure of readership. but on the other hand this distributed access (including remote photocopying) means that many papers are significantly undercounted. then there are papers whose initial submission may have had some technical processing problem, corrected by a quick replacement, and hence had many 2nd requests for a viewable version. and many people tend to use these papers as a substitute for memory -- i.e. they know a paper with a particular equation so get it just for that equation then delete it (does that count as reading?). and so on.

nonetheless some trends emerge. the bottom line paper gets about 50 requests -- we can call that the noise level. then there are the typical popular papers which get a few hundred requests, and finally there is the extremely popular ("delta function" papers that appear once every month or so, i.e. at the < .5% level, typically review or other submissions of broad cross-disciplinary interest) that instantly get a few hundred requests (i.e. in the first day after submission) and asymptote in the many hundreds or near a thousand.

finally, and perhaps most surprisingly, is that papers back to '91 (i.e. long since published) remain frequently accessed (i.e. roughly 75% of the submissions dating to '91 were accessed at least once in '94 -- and a quick check just indicated that many of these were accessed in the past two months), another indication that people find the electronic format an easier means of access than physical access to a library. (and why the vilification of high energy physicists as interested in papers at most a few nanoseconds old is so absurd).

but for the aforementioned reasons, it is dangerous to try to read too much from this data (in particular as well due to continued growth of these systems, and attendant change in habits). finally it is awkward for me to go into much greater detail since i am accessing from remote (still in french alps) til mid-sept.

Paul Ginsparg

Forward to Chapter XVI

Backward to Chapter XIV