The Blue Obelisk page about "Open Standards" says that MDL and SMILES are "proprietary format[s]" and "not Open as there is no community process for its development; it can, in principle be modified by the owner at any stage."
I have strong feelings about this topic. Why is SMILES (and I mean here the representation and not the canonicalization) a proprietary format? It was published in a journal paper, included in a book chapter, and the details are available online for many to read. There are many different implementations of the language, and some groups have added their own extensions to SMILES. Daylight and the authors of SMILES have no trademark to the term SMILES and have no patent claim. Even if Daylight were to make a new version which was not compatible to existing SMILES, almost no one would follow suit unless it proved useful.
By any definition I can come up with, SMILES is an open standard. If you count OpenSMILES as a relevant community (which is hard to justify, since nothing from OpenSMILES has made its way even into the existing open source projects) then even that objection is gone.
By comparison, several participants of Blue Obelisk say that CML is an Open Standard. (Since I can't find the license for it, I'm not sure what that means, but I've asked for that clarification as another topic.) CML is controlled by one or two people, and it seems that while the CML specification can be distributed to others, it cannot be changed without the permission of those authors.
Is that really a community process, and for that matter, what does a community process mean?
Yet another example is the MDL connection table formats. That is listed as a proprietary format, but it's also a format which is well documented by the primary maintainer, described in published papers and books, implemented in a number of code bases, including open source ones. What makes it not open, under the Blue Obelisk definition? By the way, the vendor supplied format definition should be an example of what a good specification looks like. CML's format-through-schema-dump is not.
One possibility is that the MDL specification cannot be redistributed, legally. I have a copy of it but the copyright statement prohibits sharing that spec. Is that what makes it not Open? But in that case, the ISO 8601 date format is also not an open specification, because it costs about 85 Euros to purchase and cannot be copied without permission.
If this is a problem then two ways to resolve it are 1) ask for redistribution rights, so long as there are no changes (putting it on par with CML) or 2) develop an equivalent but open documentation.
Another possibility for not calling the MDL format "open" is that they can change the format without consulting "the community." They have changed the format once, although I suspect that that involved feedback from their user community. They introduced the V2000 format, and carefully made it backwards compatible with the old file format. If they had not, others would not have followed their lead. Indeed, I haven't seen that many V2000 files in the wild. It doesn't seem like they have strong control over the format.
On the other hand, I see that the Blue Obelisk page[1] lists InChI as an open format, base solely I assume on the source code being accessible since it is a somewhat closed development process which has changed the output several times. I have looked at the source code and can say that it is not a simple specification. I am unaware of any other implementation which can parse InChI strings with near 100% success rates.
Because of the lack of a good stand-alone specification, any developer who reads the LGPL-based source code in order to reimplement it might legally be considered "dirty" in terms of clean-room coding practices, meaning that the new code will also have to be LGPL. For reference, I prefer BSD. I would have thought that a good Open Standard also means the freedom to pick a different license, including free ones which are not compatible with the GPL family of licenses.
What then are the exemplar "Open Standard" formats in this field, why are they "Open", and why are they more Open than both the MDL and SMILES formats?
ADDENDUM
I was hoping specially for feedback on which formats are open and which aren't, and why. Is SMILES open or not? Why does OpenSMILES make a difference? Are the MDL formats open? What would be needed for an OpenMDL? Is the PDB format open?
If the format is published in a journal which does not allow redistribution without paying a standard copying fee, is the format itself open or closed? For example, the SMD format published in JCICS the early 1990s was meant to be a community-driven format definition, using a context-free grammar. Is this format open or not?
