How to cite a paper whose place and status of publication is questionable or unknown

I have double checked in IEEE Xplore. The paper does not appear in the table of contents, the authors do not appear in the author list. Additionnaly, I cannot trace it in ISI Web of Science (which mentioned other conference papers by C. Kaner in 2004).

There are three options (in general):

1. A mistake in the proceedings for a regularly accepted paper,
2. The paper is more a tutorial, a keynote, a late breaking paper that has not been through the standard review process,
3. A paper that was not published in the conference, and which was put online carelessly in the conference paper shape.

As I am really unsure, and know that papers sometimes cite papers they have not read (and papers that do not exist), I would cite it only as an online document, or a preprint, with the url, without page number, and a potential note like "(often refered to as published in Metrics 2004)".

Most of all, I would suggest you not to cite it, unless it is really useful for your work. Or cite a paper really "published " by the authors instead (good luck with that).

EDIT: after a long search, I have found using the Wayback Machine that this paper could have been part of Metrics 2004 Late Breaking Papers:

The purpose of the late breaking papers session is to give authors the opportunity to present work from on-going projects, new ideas and papers not complete at the time of the original call for papers. Each paper has been peer reviewed by at least 2 independent experts.

The late breaking papers sessions allow authors to make a short presentation of their work in the main programme. Printed copies of the extended abstracts are circulated in delegate packs and, due to the timescale, the full papers will be published electronically on the conference web site.

Finally, I observe that with some online journal, it gets more difficult to have standard page numbers. Sometimes each paper gets a number. As much as I like citing papers correctly, may be page numbers are becoming useful when the work only exists in paper-like form: a DOI is an interesting alternative.