Menu
▾
▴
#299 "Mandatory when applicable" is meaningless
closed-fixed
6
2014-06-23
2011-08-30
No
The guidelines contain many instances of the phrase "mandatory when applicable" to refer to the status of an attribute. For instance, the att.internetMedia attribute @mimeType is designated "mandatory when applicable":
http://www.tei-c.org/release/doc/tei-p5-doc/en/html/ref-att.internetMedia.html
However, no-one has been able to come up with a reasonable explanation of what "mandatory when applicable" really means; it seems to be a historical survival from an early version of the guidelines, and its intent has been lost over the years. The phrase "mandatory when applicable" should be replaced with "optional" wherever it occurs in the guidelines.
Discussion
GB raised the issue of whether we should use "recommended" instead of "optional". I think this would vary from case to case, so perhaps an inventory of cases is required.
SY suggested that we aim for alignment with an existing standard terminology such as this one:
http://tools.ietf.org/html/rfc2119
which specifies the following keywords:
"MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL"
Sebastian mentioned 35 cases of the use of mwa. I find 34, one of which is in the attDef.xml file, and three of which are in <egXML> elements in valList.xml; so I count only 30 actual uses of mwa. It also shows up in TD-DocumentationElements.xml (in the guidelines text) as part of an example.
Still trying to figure out a way of generating a list of attributes which are mwa from the P5 source.
"meaningless" seems a little harsh. Earlier versions of the Guidelines make clear that the intention was to distinguish four situations :
required: the attribute or element must be supplied in all situations
mandatory when applicable: the attribute or element must be supplied in certain situations (e.g. where other elements require its presence)
recommended : good practice suggests that the attribute or element should be supplied, but it is not required
recommended when applicable : same again, but only in certain circumstances
optional : always entirely optional
However, I am quite sure that we have over the years forgotten about this distinction, and applying it retrospectively would be a nightmare. More seriously, we now have a much more robust and well thought out definition of TEI conformance which makes these distinctions unnecessary. So I have no problem with nixing all occurrences of it -- but it is NOT synonymous with "optional"!
p.s. grep is as good way as any of finding the existing cases -- just do grep mwa *.xml in the Specs directory
dont forget to remove "mwa" from even being allowed as a possible value, not just deleting current use.
Do we ever use "rwa" as a value? If so, do they all become "optional" as well?
Is it going to be safe to change all "mwa" to "recommended", or do we need to decide on a case-by-case basis which are "recommended" and which "optional"?
Oh, I see the second half of my question was already answered by Martin in the first comment below. Sorry.
Should this survey of current "mwa" uses be carried out by the Council as a whole, or is Martin happy to come up with recommendations and maybe highlight a couple of difficult cases? (Or would you like some help with that?)
At the heart of this is the problem that we have no definition for "applicable".
@Lou: Of course grep is easy, but I wanted to pull out a relatively readable collocation of all the surrounding text, so we can see what the context for each usage is.
@Gabriel: I'm happy to do the survey, but it'll take me a little while. I suspect every instance will be slightly different; some will end up being recommended, and some optional.
It's clearly not useful to retain this concept since nobody knows what it means, and it definitely has not been consistently applied. Let's nuke it.
List of uses of 'mwa' and 'rwa'
I've just uploaded a file which contains an exhaustive list of usages of 'mwa' and 'rwa' in the Guidelines. Before we make a final decision, I'd like everyone to scan through the list to see if there are actually any contexts in which either of these values makes sense. I suspect not, but it's best to be sure before we nuke it, because both values are quite widely used.
Regarding Martin's attachment: since we've all agreed that "mandatory when applicable" and "recommended when applicable" don't make sense at all, the question, I think, is whether we would want to change any cases of "mwa" or "rwa" to "rec", which seems the closest equivalent to whatever is intended by either "mwa" or "rwa". I would change @break to this for the reasons that Martin Mueller has often articulated on TEI-L: that inconsistent handling of hyphenation is the greatest barrier to interoperability he has encountered. The rest, though, strike me as no more important than any other element or attribute in the TEI, which you use if the envisioned use of your TEI documents will require that this information be machine-readable.
scanning the list, I find it hard to find any that should not be just "optional"
I'd agree with Kevin's comment of 2011-11-03.
Changes committed on SVN
When originally implemented, some instances of mwa and rwa remained in the code, specifically in attDef.xml, att.transcriptional.xml, and TD. I've now removed those, after raising a new ticket:
https://sourceforge.net/p/tei/bugs/676/
All future activity relating to this will be tracked on that ticket.