Subject: notes to netlib contributors Thank you for contributing the fruit of your labors to the general public. Here are a few points you might want to consider as you package up your work. 1. So that you can get proper credit, be sure to include your name and a journal citation as a comment in your code. If you want to get bug reports, also include an email address and date the code was last revised. Make certain that co-authors and employers are willing to have the code freely available in netlib. Add whatever copyright notice is appropriate; sample language is available by mail netlib@research.att.com send boilerplate from research. This protects both you and the user. 2. Although some large collections, like linpack, justify a separate netlib directory to themselves, most contributions are saved as a single file. For those running the Unix system, here is a shell script to combine several files into one. #!/bin/sh PATH=/bin:/usr/bin echo '# To unbundle, sh this file' for i; do echo "echo $i 1>&2" echo "sed 's/.//' >$i <<'EOF $i'" sed 's/^/-/' $i echo "EOF $i" done 3. Each contribution needs a couple sentences of description for the index. Here is an example. file a/dloess for smoothing of multivariate scattered data by Cleveland, Grosse, and Shyu ref Statistics and Computation 1:1 prec double lang Fortran77 and C alg moving least squares quadratic, k-d trees, blending functions see go/lowess To get a copy of the full schema, browse http://netlib.att.com/netlib/bib/thesaurus.html. 4. If you depend on machine constants or BLAS, please assume the user has the ones described in the netlib response to send index from blas. Otherwise, the contents of the file can be most anything you want. Avoid long lines and nonprinting characters, which are likely to be corrupted by mailers. The netlib request "send plauger_chk from research" will get you a simple C program that checks for such problems. 5. You may suggest a name you would like your file installed under; if not, we'll make up something and let you know. It will then be available world-wide from netlib@research.att.com, http://netlib.att.com/, and other netlib servers world-wide. 6. There is no refereeing. Even experimental codes may be suitable for distribution if no production-quality codes are yet available. The managing editors and, in some cases, an area editor will take at least a quick look to be sure the code seems suitable for netlib. Of course, it is in your own best interests to verify that your code is ready for public scrutiny. 7. Netlib is published from time to time on CD-ROM, by organizations (in which we have no financial interest) that will get a small profit to make it worth their while. The profit has to be small because their competition is giving it away free! If this would bother you, tell us. We think the expanded distribution that CD-ROM allows is good for everyone. Once again, thank you for participating in this free scientific exchange. If you have any questions, feel free to contact us. Eric Grosse 908-582-5828 phone AT&T Bell Labs 2T504 908-582-7415 fax Murray Hill NJ 07974 ehg@research.att.com Jack Dongarra University of Tennessee and Oak Ridge National Laboratory dongarra@cs.utk.edu