RE: Re: Maintaining a Mathematica bug list
- To: mathgroup at smc.vnet.net
- Subject: [mg97559] RE: [mg97524] Re: Maintaining a Mathematica bug list
- From: "David Park" <djmpark at comcast.net>
- Date: Sun, 15 Mar 2009 05:28:26 -0500 (EST)
- References: <14874019.1237075110231.JavaMail.root@m02>
I think this is the case of the perfect being the enemy of the useful. It is not so important that every single bug be in a maintained bug list for users. It is more important that major new bugs that affect a large number of users be on a list. I could see a useful bug list being less than a few dozen items. If some relatively savvy Mathematica user wanted to maintain such a list they might publish it on MathGroup once a week, say. They might pick up the bugs from MathGroup, or once they became known, users might send bugs to them directly. But if it is something obscure and difficult to verify the list maintainer shouldn't be expected to spend a lot of time checking it, or put it on the list. Sometimes he might just say that so-and-so reported trouble with such-and-such a function. Sometimes WRI might even send him bugs, or comment or correct the list. A short list of high impact bugs would be useful. A long exhaustive list with deep analysis and verification would be too much work. David Park djmpark at comcast.net http://home.comcast.net/~djmpark/ From: Bill Rowe [mailto:readnews at sbcglobal.net] On 3/14/09 at 5:34 AM, szhorvat at gmail.com (Szabolcs Horv=C3=A1t) wrote: >Presently the only way to learn about bugs is to follow MathGroup >closely, but even when I remember that I've seen a bug (that might >be relevant to me at the moment) mentioned a couple of months ago, I >cannot always find the posts that described it. Given Wolfram doesn't post bugs to Mathgroup, it is clear there has to be another way to learn of bugs which is to experience them yourself. The advantage of a bug list is that it (ideally) provides a means to check whether some unexpected result you get is a bug or not. But, without Wolfram's involvement I doubt this ideal will be achieved with a third party bug list. There are several issues. First, what criteria will be used to determine a given result is a bug? A new user might see Mathematica's failure to simplify Sqrt[x^2] to x as a bug. Mathgroup is full of posts from less experienced users labeling results as "bugs" more experience users know are expected and are not bugs. Assume for the moment all entries on the hypothetical bug list are truly bugs. How will the list be organized and the bugs cataloged? Given many functions in Mathematica are interdependent, a single bug could impact more than one function. Would there be a separate entry for each function affected by the bug? And for a reported bug that is validated, without the source code, how can anyone know if the bug is limited to just the specific function reported? Similarly, without the source code, how could you determine two separate reports for differing functions are really separate manifestations of the same bug? My guess is without Wolfram's involvement, the bug list will rapidly become a list of unexpected results rather than a true bug list. As a consequence, this bug list will become no more useful than Mathgroup with respect to finding information on bugs.