If we do not make it REQUIRED, then we may be looking at a lot of work coming up with use cases that generate OPTIONAL conditions.
The terms identified in RFC 2119 allow for conditions.
Parsing a sentence from STIX 2.0, 3.4 Versioning, we do assign a condition to the ‘MUST instead create’ phrase:
“If a producer other than the object creator wishes to create a new version, they
MUST instead create a new object with a new id.”
So let’s say we go with
OPTIONAL … MUST/SHALL…
These are somewhat convoluted but:
OPTIONAL – the lang: field MUST/SHALL be used in the STIX message if the producer intends to enable consumers to accelerate the language identification process.
OPTIONAL – the lang: field MUST/SHALL be used in the STIX message if the producer broadcasts to consumers who reside across a sovereign border.
Ryu asks if it’s worth spending time defining use cases?
If we don’t intend to make lang: REQUIRED, then we need to develop conditions to satisfy the business/use case and express them in the object field.
Again, that could turn into a lot of work and overly complicate the tool developer’s UI if they want to Q&A their way through the options with the user.
IMHO, tool providers can easily accommodate this field in their UI and in the interchange.
How tool providers enhance their user experience is not the CTI TC’s concern.
I believe, “REQUIRED – MUST be filled in with a valid code”, is the better choice.
7940 Jones Branch Drive, Tysons, VA 22102
Office: (703)917-7272 | Cell: (571)335-6899
From: firstname.lastname@example.org [mailto:email@example.com]
On Behalf Of Masuoka, Ryusuke
Sent: Monday, February 27, 2017 3:21 AM
To: Back, Greg <firstname.lastname@example.org>; Jason Keirstead <Jason.Keirstead@ca.ibm.com>; Allan Thomson <email@example.com>
Cc: Bret Jordan <Bret_Jordan@symantec.com>; Wunder, John A. <firstname.lastname@example.org>; email@example.com
Subject: [EXTERNAL] RE: [cti] Internationalization: lang field required or optional?
I think the differences are in use cases in each one’s
Human readable texts are for humans to consume, but
tag is for the system to produce/consume.
This is an on-the-wire/between-systems requirement/optionality.
With the system knowing the language code for the human readable
texts, the system can handle things better and provide much better UI, etc.
My question is what is worth (use cases) to define lang: tag if it is optional.
I originally didn’t feel strongly either way, but I’m coming around to feeling pretty strongly it should be optional.
Language is necessary only for human consumption (vs. encoding, which is necessary for machine consumption). IMO, fields should only be required if
leaving them off makes effective CTI sharing difficult, and I don’t (yet) think this is true for language information. It’s certainly we can specify in conformance levels or interoperability profiles, but I feel it would be a mistake to require it at the spec
As I’ve been working on python-stix2, creating an Indicator only requires “labels” and “pattern”. All other required fields (type, id, created, modified,
valid_from) can be reasonably inferred. Any program that uses python-stix2 needs to therefore require the user to enter that information, or make an assumption on their behalf. Getting the “current user’s” language works fine on personal machines, but on a
server that many people use (for example, via a web service), it’s problematic.
Also, a field doesn’t need to be required if we define how consumers should behave when it’s missing; in this case, saying that the language is “undefined”
or “unspecified” is likely OK, particularly that “unspecified” is OK for machine-to-machine communication that doesn’t involve humans. This is the reason I’ve always felt “modified” should be optional; IMO it’s perfectly reasonable to mandate that, if not
explicitly specified in JSON, consumers MUST assume it was last modified at the “created” date.
I also agree with Alan and John in the preference to make this optional.
In general I do not like sending bytes when bytes are not required in a data interchange format, especially when considering the scale of data we will be dealing with
in STIX/TAXII. We should be looking for opportunities to keep the data format trim. Truthfully, the vast majority of data in an ecosystem will all be the same language, and thus having to transmit a language tag for every single object in a package is redundant
There is also another issue with making it "required", and that is that we would then have to support "unknown" or "undefined" - which many products would have to mark
content as since they may not know the producer of the content's native language. There is an ISO 639 language tag for "undefined", but there is no IETF tag for "undefined" in the IANA registry, they never adopted the ISO entry. So making this mandatory may
force a revisit of the RFC5646decision.
STSM, Product Architect, Security Intelligence, IBM Security Systems
Without data, all you are is just another person with an opinion - Unknown
From: Allan Thomson <firstname.lastname@example.org>
To: Bret Jordan <Bret_Jordan@symantec.com>,
"Wunder, John A." <email@example.com>, "firstname.lastname@example.org" <email@example.com>
Date: 02/23/2017 07:01 PM
Subject: Re: [cti] Internationalization: lang field
required or optional?
Sent by: <firstname.lastname@example.org>
If you are expecting to use different language content then its required for interoperability reasons.
But by marking it required in the spec means that all content must have it even when most content is not multi-language.
I generally would prefer more tolerance in the spec level and let the products/market use good behavior to drive what fields are included or not.
If people care about language and multi-language support then they will use it. If they don’t then they wont be interoperable as that will be part of the test in
the interop spec.
Bret Jordan <Bret_Jordan@symantec.com>
Date: Thursday, February 23, 2017 at 2:04 PM
To: Allan Thomson <email@example.com>, "Wunder, John" <firstname.lastname@example.org>, "email@example.com" <firstname.lastname@example.org>
Subject: Re: [cti] Internationalization: lang field required or optional?
1) In reality we are talking about a feature not a property.
2) If it is property of this feature is optional, then the only products that will implement this feature, are those that care about internationalization.
3) If it is required, then everyone will be forced to implement it.
Personally I see this as a data quality issue, not a STIX issue. And I think both sides can suffer from it.
Problems with Required:
a) product or tool does not care, does not provide a UX for it, and just hard codes it to something, say "en"
b) product or tool does provide a UX for it, but analyst does not care and it just remains what ever the default is.
Problems with Optional:
a) product or tool does not care, does not provide a UX for it, and just leaves it out of the data. So it is undef.
b) product or tool does care and provides a UX for it and the analyst does not care and leaves it blank.
c) Broker product or tool takes in data that has a lang tag, but they do not support that feature so they never implemented it. So when the data goes back out the other side, the
language tag is now missing.
I personally do not see the harm in requiring tools to support and populate the Lang tag. In the spec we can define an "unknown" value, so if you are doing bulk loading of data and
you honestly do not know the language, you could just flag it as "unknown". Then at least as the consumer you would know that the producer did not know the language. Versus getting an object where the language tag is omitted and you do not know if:
i) they did not know the language
ii) there tool did not support it
iii) they were just lazy and did not add it.
Once again, this is a data quality problem and if we make the lang field required, then it is a SUPER EASY interop test to see if they do it right. If it is optional, then you are
just at a guess all the time.
email@example.com <firstname.lastname@example.org> on behalf of Allan Thomson <email@example.com>
Sent: Thursday, February 23, 2017 2:29:59 PM
To: Wunder, John A.; firstname.lastname@example.org
Subject: Re: [cti] Internationalization: lang field required or optional?
"email@example.com" <firstname.lastname@example.org> on behalf of "Wunder, John" <email@example.com>
Date: Thursday, February 23, 2017 at 12:59 PM
To: "firstname.lastname@example.org" <email@example.com>
Subject: [cti] Internationalization: lang field required or optional?
We’re getting very close to having a completed approach for internationalization, you can see the full writeup here:
We do have one remaining question before we can move forward though. As part of the proposal, every single top-level object has a “lang” field, that identifies the
language of the text content in that object. What we need to decide is whether we make that field required or optional.
If we make the field required, every top-level object in STIX (SDOs and SROs) would have to have a “lang” field in it or it would be invalid STIX. If we make it optional,
producers could either include the field or not.
Here are some thoughts:
Making it required:
- All SDOs and SROs would have a language tag, so consumers could depend on
it being there
- It would encourage producers to actually fill it out, because they wouldn’t
be creating valid STIX otherwise
- It shows we have a commitment to internationalization
Making it optional:
- Any SRO or SDO could have a language tag, so consumers could not depend on
- Producers would not have to create it
- We do have a SHOULD requirement saying that it should be included
My opinion is that we should make it optional. If it’s required, I think people who don’t want to do internationalization (especially those creating one-off scripts
or open source tools) will hardcode it to English and things will be mislabeled. If it’s optional, I think those who need/want to support internationalization and would do it right (most/all vendors, major open source projects) will populate it correctly regardless…because
they need it…while those who couldn’t be bothered will be able to leave it off and we won’t have mis-labeled data. Also it’s almost not worth saying, but we already have a bunch of required fields on every SDO/SRO and I’ve already had one conversation with
someone who said there’s a lot of bloat…would like to avoid adding to that.
Anyway, what does everyone think…required or optional?