Short Answer: dblp is still pretty incomplete, and will probably ever stay so.
The dblp team strives to provide an as comprehensive as humanly possible computer science bibliography, but true completeness can (of course) never be reached. We havea rather small team of editors
and a pretty rigorous process of quality checking all new additions to dblp, and each day a huge amount of new scholarly works is released to the public. Hence, our resources are limited, and so we have to inevitably select and prioritize some possible additions over another.
There are usually three major factors that are limiting the comprehensiveness of dblp:
- dblp is a bibliography focused on topics from computer science. However, where a field of research starts or where it ends is up to debate, and a lot of scholarly work takes place in interdisciplinary or hybrid fields. Hence, since we have to draw a line somewhere, interdisciplinary fields will probably never be covered as comprehensively as we aim to do with core computer science fields.
- But even just in core computer science, the vast amount of scholarly works is too much for a small team and a quality focused workflow like ours. Hence we aim to tackle our task in a top down approach by assessing which volumes and issues are of the most interest to the general, international computer science community. This decisions are guided by the criteria and principles defined by our advisory board. But they are often tough calls, and made case by case.
- Finally, but maybe even most importantly, the availability of open bibliographic metadata (or, rather, the lack thereof) is always limiting factor for our indexing decisions. We cannot index what we cannot find and reproduce from reliable, open sources. Although the situation has improved in the recent years, this is still a limitation for a remarkable large fraction of all indexing requests we receive.
Although we ourselves do not have any metric or study to answer this question with scientific rigor, there exist a number of scholarly articles that discuss the issue of the completeness of dblp. However, those studies are several years old, and a lot has improved with regard to our coverage during the recent years. (If you know of any more recent studies, we'd appreciate if you could drop us an email.) But if you are looking for a study of this topic, you might want to start looking here:
This paper examines the difference and similarities between the two on-line computer science citation databases DBLP and CiteSeer. [...] The model permits us to predict that the DBLP database covers approximately 24% of the entire literature of Computer Science.
Our results show a total coverage of 65% at the end of 2005 for the conferences listed by Laender et al. and Martins et al. which is only a small subset of all computer science conferences. However, by the way this list was created, we can assume that is contains the most relevant ones.