--------------------John Ewing's Reply--------------------------- From JHE@MATH.AMS.ORG Wed Jan 15 15:57:35 1997 Received: from axp14.ams.org (AXP14.AMS.ORG [130.44.1.14]) by euclid.math.temple.edu (8.7.4/8.6.12) with ESMTP id PAA03400 for ; Wed, 15 Jan 1997 15:57:34 GMT Posted-Date: Wed, 15 Jan 1997 15:57:34 GMT Received-Date: Wed, 15 Jan 1997 15:57:34 GMT Received: from AXP14.AMS.ORG by AXP14.AMS.ORG (PMDF V5.0-6 #16534) id <01IE8MT98WC00008T1@AXP14.AMS.ORG>; Wed, 15 Jan 1997 10:56:59 -0500 (EST) Date: Wed, 15 Jan 1997 10:56:59 -0500 (EST) From: "John Ewing, Exec Dir" Subject: Re: Please save the honor of Shalosh B. Ekhad In-reply-to: <199701131347.IAA21775@euclid.math.temple.edu> To: Doron Zeilberger Cc: JHE@MATH.AMS.ORG, rkd@MATH.AMS.ORG, jek@MATH.AMS.ORG, dgb@MATH.AMS.ORG Message-id: <853343819.715294.JHE@MATH.AMS.ORG> MIME-version: 1.0 Content-type: TEXT/PLAIN; CHARSET=US-ASCII Content-transfer-encoding: 7BIT Mail-system-version: Status: RO Content-Length: 4966 Dear Doron: This one may get complicated. Here goes. First, we all COMPLETELY agree that in an electronic age, errata ought to go up as soon as possible, and be immediately apparent to any reader at the location of the error. So when mistakes are discovered in MathSciNet --- the web version of Math Reviews --- it is our policy to immediately make the correction and post it, side-by-side with the original mistake (which is kept for the sake of good scholarship). Here is the entry in MathSciNet for the Ekhad review you mentioned: TEXT OF REVIEW ON MATHSCINET: ************************************ Dixon's theorem (as stated) provides a symmetric-type series representation for the multinomial coefficient $(n+b+c;n,b,c)=R(n)$. The author attempts to give a very short and very elementary proof by means of the principle of mathematical induction. After introducing some special symbols, he states "since the theorem is obviously true for $n=0$, it would follow by induction once we show that $(n+1)R(n+1)- (n+b+c+1)R(n) \equiv0$". But this identity is an immediate consequence of the definition for $R(n)$, and hence the "proof" lacks validity. ERRATUM: In this paper $R(n)$ is defined to be the sum on the left-hand side of Dixon's identity and not, as is stated in the review, the multinomial coefficient. The method and proof used in the paper are valid. Reviewed by Gloria Olive {For errata and/or addenda to the original MR item see MR 91m Errata and Addenda in the paper version} ************************************ And for all mathematicians who access MathSciNet, this is exactly what they see. The CONFUSION arises because we have offered Math Reviews in several formats, some of which appear to be web services but in reality are not. Of course, we have offered Math Reviews on compact disk, and we still do offer it that way. Making corrections on a disk are essentially as difficult at making them on paper; we can only publish the errata and point to the mistake after it has been made. More difficult is the case of DIALOG, which is an electronic service to which we give the right to sell information from our database. Dialog is NOT MathSciNet. Indeed, it is not a part of the AMS in any sense, but rather a service provided to libraries and scholars. We provide the Math Reviews database to them on tape; they load it onto their computers; and we provide regular updates --- much the same as one provides paper volumes of Math Reviews to a library. They can ADD to their database, but not easily modify it. At the moment, in order to change a piece of data in that database, the entire database must be reloaded --- which requires many hours and a great deal of work. We have discussed ways in which to modify the database, but this would require a large and expensive effort, that ultimately would have to be agreed to by both Dialog and the AMS. For MathSciNet (http://e-math.ams.org/msnprhtml/review_search.html), which is available over the web, and provided from our own web servers, we have complete control. We put in place a procedure for changing the database when corrections are necessary. That procedure does not require reloading the entire database (a project that requires nearly a day, by the way), and it is a part of the normal daily updates. Here is a summary: For paper, CD-Rom, and Dialog, we have no good way to make corrections to the data so that users are immediately aware of the corrections. Corrections are made, of course, but not in a way that makes them clear AT THE POINT OF ERROR. For MathSciNet (our web version), it is our policy to make corrections immediately and visibly, usually on the day after they are discovered. The fact is that MOST users of the electronic version of Math Reviews use MathSciNet; only a small number use Dialog. Indeed, the problem you mention is one reason we are not happy about the continued use of Dialog, but it serves a small segment of the community and so we believe it is better to leave it in place until we find a better alternative. In the meantime, although the problem of not being able to make corrections affects only a small number of users, I agree that it is a serious one. We'd like to be able to make corrections on ALL versions of Math Reviews in the same way we make them on MathSciNet, and we'll keep looking from better ways to do so. Of course, the problem is solved if everyone switches to MathSciNet --- which is a pretty good product, and which is spreading rapidly. Complicated? Sure . . . and not entirely satisfactory, I know. But I'd say that 95% of your solution to the problem is already implemented, as a regular feature of MathSciNet. We'll keep thinking about the other 5%, however, and try to find a cost effective way to implement it. John ----------------------------- John Ewing, Amer Math Soc ewing@ams.org 401-455-4100 ----------------------------- ----------------------end reply----------------------------------------------