More info on Brazilian e-voting

Brazil's e-voting system is widely touted by Brazilian officials, and others, as the answer to e-voting problems. PR material is produced in English by the Brazilian authorities to sing its praises.

Academics in Brazil, however, have for some time urged caution and have produced their own - less publicised - versions of the story. And now, as someone who seems to be on the 'usual suspects' lists when it comes to e-voting, I've just received a link to a formal expert report on the Brazilian system, from someone who has posted the material to the Web, but prefers to remain anonymous (a former Electoral Justice dept. worker in Brazil with access to such documents).

One item is a 2002 report commissioned by the then Brazilian opposition from independent experts, "that shows the poor software development made by the TSE and Unisys". The text is in Portuguese.

Here's my own short summary of the report - my Portuguese is not perfect, but I've been told by Brazilian e-voting experts that it's a fair reflection of the contents:

The report is a quality assessment, covering documentation and code, and was commissioned by the Partido dos Trabalhadores (Workers' Party), via the COPPETEC Foundation, and produced in August 2002.
The assessors were Professor Ana Regina Cavalcanti da Rocha, Professor Guilherme Horta Travassos, and experts Gleison Santos de Souza and Somulo Nogueira Mafra.
The results are divided into (a) software documentation and (b) code.
The document contains general comments concerning software documentation - details are contained in Annex 4, not supplied.

Software documentation was divided into two groups. The first group was found to be based on "in some respects out-of-date and incoherent methodology"; specifications were "inadequate, incomplete and inconsistent"; poor project definition; no documentation on quality procedures, configuration management procedures or systematic testing procedures (meaning either no action or no documentation of it); no test plan documentation; when further test planning and results documentation was requested, this was incomplete with results obtained missing.

For the second group of documents, seemingly primarily the responsibility of Unisys, a variety of inadequacies are reported, including lack of documentation of differences between the test and live systems, and of modifications to code. While FTP file transfer is mentioned, there is no documentation of security measures with respect to FTP.

The report points out that [the presence of] "processes for developing quality software are indicators of products of good quality" and goes on to conclude that the software development process was "substantially ad hoc and immature" which "often leads to products of unpredictable quality, strongly dependent on the characterisitcs of the developers".

In the second section, concerning code, the report states that there are sections of inactive code in several applications with no explanation for their presence; evidence of poor project management with different solutions found by different programmers for similar problems; lack of treatment of error conditions...etc. The report also remarks on the lack of a watermark on the voting paper trail.

The report concludes with a set of short-term recommendations, and longer-term recommendations.

There are many other points, and anyone with decent Portuguese should feel free to render a fuller summary, for the benefit of those outside Brazil.