Dear Australian Electoral Commission,

This request is in two parts. I am happy for you to consider the request as a single request as they are all closely related, or as multiple requests if that is preferable for you.

* Part 1: source code for counting software

I am seeking source code for the software used to conduct the count of votes for a Senate election.

This request includes scripts or interpreted code used within another piece of software (for example, T-SQL scripts, stored procedures etc).

This request excludes software used for data entry or for interpretation of those scripts but includes data validation software if that is distinct from data entry software.

This request may encompass more than one piece of software and I seek source code for each.

* Part 2: data specifications

I am seeking any documents which describe bespoke data formats used by any of the software sought in Part 1, either as input or output formats.

This request excludes any data formats which are human readable or for which published specifications are available (e.g. PDF).

For clarity, the types of documents I am seeking may include database table specifications, EBNF specifications for bespoke input data, column descriptors for CSV files, XML schemas or similar documents.

I am seeking a waiver of all associated access charges on the basis
that publication of these documents would be in the public
interest. In particular it would provide public confidence that the electronic count is conducted in accordance with the law.

Yours faithfully,

Michael Cordover

Dear Australian Electoral Commission,

Please pass this on to the person who conducts Freedom of Information reviews.

I am writing to request an internal review of Australian Electoral Commission's handling of my FOI request 'Software by which Senate counts are conducted'.

A full history of my FOI request and all correspondence is available on the internet at this address: https://www.righttoknow.org.au/request/s...

I also recognise that things are quite busy at the AEC at the moment. I am happy for this to be taken to have been received as late as 3 December (30 days from when the original decision was made) and therefore not to receive a response until early January. However I am keen not to prejudice any external review rights that might arise and I consent to the increased time on that basis.

Decisions to be reviewed
------------------------

I refer to the statement of reasons for refusing access given by Paul Pirani dated 4 November 2013 (received 5 November 2013). There are three decisions contained in this letter for which I seek review:

1. The decision not to disclose the schedule of documents (reasons at [24])

2. The decision that all of the documents are exempt under s 47(1)(a) as trade secrets (reasons at [7])

3. The decision that all of the documents are exempt under s 47(1)(b) as containing commercially valuable information, the value of which could reasonably be expected to be destroyed or diminished by disclosure (reasons at [7])

As they are closely related, I will deal with the second and third decisions together.

Matters of policy
-----------------

Before particularising the reasons I believe the decision was wrong at law, I note the following matters which suggest to me that the AEC should release the documents I request as a matter of policy.

The AEC s 9 FOI statement (http://www.aec.gov.au/about_aec/Publicat...) provides that the Easycount Senate User Guide is available under FOI and does not indicate that it would be exempt or partially exempt.

The AEC indicated in its supplementary submission (number 181, dated 7 February 2003) to the JSCEM inquiry into the 2001 Federal Election at [8.12] that:

"In the interests of transparency, and because there are no security implications, the code [for EasyCount] will be available for review."

As that statement recognises, confidence in the electoral system can only exist where the system is transparent. The AEC does an exceptionally good job and I do not suggest any improper motive for refusing to disclose this material. However, despite s 273A(5) of the CEA, to withhold this information is inconsistent with the the general openness of the AEC's dealings.

I recognise, of course, that those policy reasons are not sufficient to suggest the original decision be varied and that there is no public interest test in the s 47 exemptions under the FOI Act. For that reason, other than as disclosed below, I do not rely on those policy reasons in seeking a review of the decision.

Schedule of documents
---------------------

The reasons at [24] state that disclosure of the schedule would "give general guidance to a person on how to uncover the trade secret protecting the EasyCount Software."

This is predicated on there being a protected trade secret in the EasyCount software. I will explain below why I do not believe that to be the case.

However, even if there is a trade secret which makes the software source code exempt, I do not believe the list of documents also attracts that exemption.

Mr Pirani relies on an exemption under s 26(2) of the FOI Act. This exemption applies where the schedule would be an exempt document. Given the wording at [24], the reasons clearly imply exemption is claimed under s 47(1)(a) as the schedule would disclose a trade secret. However s 47(1)(a) requires disclosure of a trade secret; that it would "provide guidance ... on how to uncover [a] trade secret" is not sufficient. This type of material is perhaps more analogous to "know how" which is not protected.

To be protected the schedule must be itself of commercial value and confidential in nature. At most the schedule discloses the functionality and structure of the EasyCount software and its documentation. It does not disclose the way in which this functionality is implemented. This amounts merely to a statement of purpose, not to information which is of the type which receives the protection of confidentiality.

Furthermore that meta-data is already the subject of disclosure by the AEC. Manuals for Senate and fee-for-service election editions of EasyCount are listed on http://www.aec.gov.au/about_aec/Publicat... as being available by FOI; this at least discloses their existence in the same way as the schedule would. Significant details about the structure and functionality of EasyCount are in the public domain, having been disclosed in the AEC's supplementary submission dated 7 Feb 2003 to the JSCEM enquiry into the 2001 election (submission no 181).

Even if some elements of the schedule would be exempt, certainly not all of the schedule is exempt. The names and nature of some documents have already been published. In those circumstances the AEC should provide at least an edited copy of the schedule under s 22 of the FOI Act.

Trade secrets & commercially valuable information
-------------------------------------------------

I accept the definition of trade secret given in DEWRSB v Staff Development and Training Company (2001) 114 FCR 301 and reiterated in the OAIC Guidelines on FOI required to be taken into account by s 93A of the FOI Act.

To be a trade secret, the information must be able to be put to advantageous use by someone involved in an identifiable trade (DEWRSB at [43]). The decision identified two areas of competition at [18]: industrial elections (under the Fair Work (Registered Organisations) Act) and fee for service elections.

My request was not for documents relating to those elections. My request was solely for documents relating to the senate count. Consistent with the reasons, this is not subject to any degree of competition.

The decision relies on the claim at [14] that the code base for EasyCount is shared between editions to such an extent that the fee for service versions are inseparable from the senate count versions. However, at [18](c) it is made clear that both industrial and fee for service elections have customised versions of EasyCount.

Furthermore the different counting mechanisms must form separate subroutines or functions within the computer code (if they did not, the counting method would be the same). As such those parts of the code are necessarily separable.

The decision clearly makes no attempt to provide an edited version of the documents under s 22 of the FOI Act. On the basis that the senate count functionality is separable, this is a clear error in law.

In addition, however, I contend that there is no trade secret even in the versions of EasyCount used for fee for service and industrial elections. In essence my position is that this material has no commercial value, or that the commercial value would not be diminished by its publication, or that the any advantage the AEC holds would not be diminished by publication. This is sufficient to deal with both claimed exemptions under s 47 of the FOI Act.

To explain my position it is necessary to differentiate the source code of a program (the instructions in a particular programming language) and the algorithm used by the program (the generic description of the way in which a result is achieved). Disclosure of the source code results in disclosure of the algorithm. However algorithms can be implemented in other languages once known.

As an example, the bubble sort algorithm is a well-known way of sorting a list of data. The algorithm is a way of doing things (an idea) which can be implemented in a range of programming languages.

The source code of computer programs is protected by the Copyright Act. The disclosure of source code under the FOI Act does not limit the applicability of copyright. Using or making copies of the source code, or creating an adaptation derived from the source code, would be an unlawful infringement of copyright.

On that basis no competitor could lawfully use any source code released under the FOI Act. For that reason the commercial value of the source code would be preserved and the advantage the AEC holds would not be diminished. For that reason the source code itself cannot constitute a trade secret or attract the protection of s 47(1)(b).

Algorithms do not attract the protection of copyright. Therefore an algorithm could attract the protection of s 47(1). To the extent that these algorithms are trade secrets (or commercially valuable) the release of source code may result in their disclosure and a loss of commercial advantage.

In order for these to be capable of being trade secrets they must be confidential. The algorithms used by EasyCount are not confidential. The algorithms used for various forms of industrial elections are all specified at http://www.aec.gov.au/About_AEC/AEC_Serv... with sufficient detail to re-implement them. The Senate and House of Representatives electoral count algorithms are described at http://www.aec.gov.au/Voting/counting/in... and in the Commonwealth Electoral Act.

Any system implemented in EasyCount which is not so described is likely described algorithmically elsewhere.

Any system algorithmically described could be reproduced easily by any programmer. In accordance with Dais Studio Pty Ltd v Bullet Creative Pty Ltd [2007] FCA 2054 at [77]-[80] this is a relevant consideration to whether the material is capable of being a trade secret. Where it is easily reproduced it is not capable of being a trade secret. This operates in addition to the fact that the algorithms are not confidential.

In particular, because the algorithms are already known, their disclosure by the AEC cannot result in a loss of commercial value. They have no commercial value because they are not secret.

The only algorithms which could constitute trade secrets are those which count votes in a unique way that has never otherwise been publicly disclosed. I accept that such an algorithm could constitute a trade secret to the extent it was used in industrial or fee-for-service elections. However, if the counting method is broadly known, for example having been disclosed to a wide range of electors, this would diminish the degree of protection available.

Identical arguments apply to the disclosure of information about data structures representing votes and documentation describing the operation of the software.

To the extent that any of this argument fails, relevant portions of the documents should be excluded and an edited version of the requested documents provided under s 22 of the FOI Act.

Should you wish for any clarification or to discuss this please do not hesitate to contact me by reply email.

Yours faithfully,

Michael Cordover

Dear Owen,

Thank you for this response.

Would it be possible to get unsigned copies of each letter as a word document or PDF, so I can copy and paste from them?

If you'd prefer to send those copies to a private email address please let me know.

Many thanks

Michael Cordover

Dear Owen,

I agree that you have no further obligations under the FOI Act in relation to this request.

I was hoping, however, that as a matter of courtesy you might provide me with the letters in a more usable format.

I note that I could perhaps make another FOI request for those response letters in their settled but unsigned form. If I were to request access in the form of a digital copy of the Microsoft Word files I suspect s 20(2) would then require they be provided to me in that form.

I will not make such a request, however, and I will not push this matter further.

Thank you again for your assistance with my requests.

Michael Cordover