Single Patterns Criteria Protocol

• no single form is imposed but consider using a known and proven form
• if you use your own form, clearly motivate it and make sure that people don't get lost

• demonstrate pattern character by showing applications
• best: refer to published designs
• second best: refer to unpublished designs/projects from other people
• third best: refer only to your own projects/designs
• forth best (actually, pretty bad): no known uses

• try to put your pattern in context. which related/similar patterns exist? this will help us editors a lot and will serve the community as a whole since classification efforts will be undertaken.

• The paper should not over-theorise.
• The pattern should stand for itself i.e. several people should express that they have learned something from reading it e.g. "yes, I do that and it works" or "that's great, I can use that". Remember patterns are for sharing insight.
• Most theoretical discussions of patterns that I have seen (i.e. discussions that don't proceed by proposing specific improvements) require more experience with code (aka designs ...) to resolve, not more discussion.

(regarding pattern form)

do we have any proven forms? i'm sure we do, but it would be interesting to discuss how we know they're proven. i'd be surprised if we agree. provacative starting point: i do not consider the GOF form to be a proven form. note that this is a little different from saying whether the application of the patterns are proven, but the two are very closely linked. Alexander said a pattern is a "thing, and a description of the thing," and that both the thing AND its description should have the QWAN. that means, to me, that they must work as literature.

JimCoplien, June 1996.

(regarding pattern form)

Point taken. I wanted to help people consider both form and content as an important issue. Many pattern description formats just seem arbitrary and confusing to me. If people don't have a clear picture of their form they are probably better off sticking to a well-known form like the GOF one into which already considerable thinking has been invested.

DirkRiehle, June 1996.

(regarding "used twice" rule)

I support this idea although there will be people who think differently about the order: "A good design idea should get its chance to be published, even if you can refer only to your own designs." I had a discussion about this in my homegroup at last year's PLoP. There was no common agreement.

FrankBuschmann, June 1996.

(regarding "used twice" rule)

I STRONGLY feel that we accept only those papers that refer to more than two existing designs. I am less concerned that the designs be published, but authors should strive for those when possible. (I had to do a lot of digging last year to unearth decades-old documents that documented the patterns submitted by the AT&T switching folks...)

Let me motivate my position with some lighthearded cites.

This first example may not make it with the Europeans (certainly not with Bruce :-)) but:

I have but one lamp by which my feet are guided, and that is the lamp of experience. -- Patrick Henry (an American revolutionary patriot)

In any case, I think this principle stands as the largest foundation, if not the sole foundation, for the credibility of patterns. There is a lot of industry research that suggests that you should find three prior uses:

• Rules of Three (Biggerstaff, T., and C. Richter. Reusability framework, assessment, and directions. Hawaii International Conference on System Sciences, January 7-10, 1987, pp. 502-512.)

• There's lots of related stuff by Will Tracz (his talks at the Three Rs conference, don't have the cite off-hand).

• Jerry Weinberg: "Once is an event; twice is a coincidence; three times is a pattern." (He was saying this long before we were saying the word "pattern"; by the way, he's a big fan of what we're doing.)

JimCoplien, June 1996.

(regarding: try to put your pattern in context. which related/similar patterns exist?)

This is a very hard job that requires a lot of time. It would be great, but I don't think that this is possible. We spent a lot of time to connect the patterns of our book with the rest of the pattern universe. But I think there are still a lot of patterns to which we can tie ours, but we didn't know about these patterns or we didn't see the relationship clearly enough.

FrankBuschmann, June 1996.

(regarding: try to put your pattern in context. which related/similar patterns exist?)

Thus, we should welcome such an effort, but it should not be a must.I think this is becoming more and more important. At PLoP/95, I saw a few authors who hadn't done their homework, and their patterns were minor variations on existing patterns. Their patterns should have appeared as footnotes to the original ones. Sometimes, they didn't even cite the original ones.

I would guess that, in the domain of reliable telecommunication software, there may be only 100 patterns. Heck, CA only had 253 patterns that covered everything from countries to pictures on the wall, and we shouldn't be so proud to think our discipline is that much broader. Yet, we have probably unearthed over 200 telecommunication goodies that we call "patterns", most of which stand alone. I think it's a shame. We should be careful to describe which are variants of existing patterns. We should be careful to describe how patterns work together; that will lead to pattern languages.

If we don't do this, we'll end up with dozens of books on patterns, with incredible duplication, redundancy, and contradiction between them. We can clean it up some day, but I think some day must be now. If we don't do it, the literature will fall into disuse. It will collapse under its own weight.

Maybe we should give people the index that came out of the meeting Gerard hosted, and ask people to tie their patterns into that framework (poor word choice, perhaps, but...).

(This was our plan. Unfortunately, I have been moving slowly on it. RalphJohnson)

If you write a philosophical treatise, there is more work in researching the writings of those who have gone before than there is to write the newly captured idea. We should be the same.

More to the point, we should encourage authors to take this on themselves. I think it's a matter of adhering to the minimal standards of scholarship. And of good literature. And that's what we're about. The authors should try; they may not succeed in seeing all the tie-ins, but that's O.K. Some things can be left to time. But I feel we should require, now, that they start making the effort with the existing pattern literature.

JimCoplien, June 1996.

(regarding the last paragraphs by Jim Coplien)

Currently people are putting out their patterns. We clearly should start consolidating soon.

What about OBST, the "Observer Task Force?" (Obst is fruit in German (and a research OODBMS)--three uses, we have a pattern here!!)

People interested in a specific (more fundamental) pattern might already team up and try to get a grip on its variations for a Best of PLoP or the like.

DirkRiehle, June 1996.

(regarding the last paragraphs by Jim Coplien)

I agree completely, and whole-heartedly. I think this is something we should emphasize at PLoP and that the book editorial board should also emphasize. People should not only reference the pattern literature, they should reference the non-pattern literature. This is harder, but also important. It will also be hard to make citations without messing up the pattern. Nevertheless, it should be done.

RalphJohnson, June 1996.