Bugs, Blanks and Errors

... the names are all wrong...

When Prof. April sat for the first time in front of FishBase, he decided to call up a group of South American killifish, and after a quick look through the list of species, he informed the astonished FishBase Team that "the names are all wrong." When we followed up and asked for a reference, it turned out that Ms. May, a student of Prof. April, had recently completed a thesis that strongly modified the taxonomy of this group and largely disagreed with a previous revision, the basis for the information in FishBase.

Though the names in the above story (though not the story itself) are fictional, it serves to illustrate a number of issues concerning the quality of information in FishBase. Most first-time users and, unfortunately quite a few of those who wrote reviews of FishBase (see the ‘Making of FishBase’, this vol.) tend to look up the species that they know best. Not surprisingly, as would occur with any encyclopedia, they then find that they know something about these species that FishBase does not know. In contrast to a printed encyclopedia, however, they can supply the relevant reprint to the FishBase Team. They will then find their species well covered in the next update, and their name in the list of people who helped improve FishBase. However, even before this happens they will usually find a new piece of information about the ten species they know best. And of course, they will also find relevant information on the 24,990 species with which they are not familiar.

In the case of Prof. April, the situation was more complicated because Ms. May’s thesis had not yet been exposed to the judgment of other taxonomists, who might decide to ignore the taxonomic re-arrangement proposed therein. However, this is not an attempt to belittle the presence of bugs, blanks and errors in FishBase, but rather to serve as introduction on how we deal with them.

A work of size inevitably contains errors

A work of this size and complexity will inevitably contain errors and discrepancies. The problems that users of FishBase are likely to encounter are of four basic types, presented here in descending frequency of occurrence:

i) Empty fields, though information does exist that could have been used;

ii) Erroneous entries, i.e., either entries not supported by the cited reference, or reproducing a manifest error in that cited reference;

iii) ‘Bugs’, i.e., routines that do not perform the functions they were designed to (Myers 1979; Bruce 1980; Ozkarahan 1990; Pfleeger 1992); and

iv) Tables that should have been designed differently.

We adjust tables to fit the data

To deal first with item (iv); we propose you read in this volume the background to the table in question. If you still think that it should be redesigned, please contact us, let us know of your reasons, and the data which support them. We will very likely adjust the table to fit the data.

Empty fields (i) are a ‘problem’, and we are doing our best to fill as many fields as possible for as many species as possible. However, the information required may not have been published or may not be available to us, or we may not yet have had the time to use a publication completely. Please send us any publication which you think would be useful for filling in a field or table that would otherwise remain blank (see ‘How to Become a FishBase Collaborator . . . and Why’, this vol.). Collaborators in countries, provinces or projects who want us to focus on the species of their respective areas are invited to consider supporting us with modest funds (as done by Mexico, British Columbia, Australia and MRAG and considered by New Zealand and the Mekong Secretariat) or sending personnel to work with the FishBase Team (as done by Taiwan and, indirectly, by Denmark, The Netherlands and Finland).

Send us an email

FishBase differs from numerous other databases, especially those created by individual researchers, in that it is widely available. This implies that the errors in (ii) above are exposed to the critical gaze of a large numbers of users. Some may scoff, and perhaps dismiss as unattainable our attempt to present reliable key information on all fish species of the world (e.g., Turner 1997). Other users¾ and we hope they will be a majority¾ will send us an e-mail to point out our errors (or those of our sources), and provide alternative entries (and/or sources). We believe that if this happens, most of egregious type (ii) errors will be purged from subsequent versions of FishBase.

Two months of bug-hunting found a lot

The type of problems FishBase users may encounter in (iii) are the true bugs of our title. The following step-by-step procedure was followed (by M.L.D. Palomares) to reduce the number of bugs to a minimum:

a) For all tables, verify that:

  • all links are properly connected, i.e., that buttons opening other windows do open the designated window and that buttons running routines or graphs, e.g., Print buttons, do run the designated routines.

b) For all fields of a table, verify that:
  • choice fields consist of mutually exclusive and comprehensive choices;

  • fields linked to other windows, e.g., reference fields, are properly connected; and

  • field values that are automatically computed by internal routines are numerically correct.

c) For all procedures, verify that:

  • the buttons accurately run the designated procedures or graphs.

The extensive list of bugs caught in this annual process is handed over to the persons responsible for the tables in question and to FishBase programmer, Ma. Josephine France Rius for FishBase 2000. Then, the last step was to:

d) Verify that all bugs spotted in (a) to (c) were fixed.

This process was also used to ensure that all forms of graphs and reports followed an agreed standard, that screen prompts were straightforward and easily understood, and that all technical terms were covered in the GLOSSARY.

FishBase, being as large as it is, precludes us from guaranteeing that this procedure picked up all bugs. However, we do guarantee that we will repair any remaining bug you bring to our attention.

Acknowledgment

We thank all FishBase users, past and future who contributed (or will contribute) to making FishBase as free of bugs, blanks and errors as possible given our means.

References

Bruce, R.C. 1980. Software debugging for microcomputers. Reston Publishing Company, Inc. Reston, Virginia, USA. 351 p.

Myers, G.J. 1979. The art of software testing. John Wiley and Sons, Inc., New York. 177 p.

Ozkarahan, E. 1990. Database management: concepts, design, and practice. Prentice-Hall International, Inc., New Jersey, USA. 560 p.

Pfleeger, S.L. 1992. Software engineering: the production quality of software. 2nd ed. St. Martin Publications, Quezon City, Philippines. 517 p.

Turner, G. 1997. Book reviews: FishBase 96: concepts, design and data sources. Rev. Fish Biol. Fish. 7(3):374-375.

Maria Lourdes D. Palomares, Rainer Froese and Daniel Pauly