Of Layer Names — And Babies — And Bathwater

Harrison Eiteljorg, II

(See email contacts page for the author's email address.)

In the last issue of the CSA Newsletter was an article about naming layers in a CAD system — Gregory Tucker and John Wallrodt, "Rethinking CAD Data Structures — Pompeii Archaeological Research Project: Porta Stabia," XXVI, 1; April, 2013;csanet.org/newsletter/spring13/nls1302.html. The article set forth the system used to name and access layers in AutoCAD for the PARP:PS project.

The authors referenced my work about this, the CSA Layer-Naming Convention, and a modification of it discussed in a CSA Newsletter article — Paul Blomerus and Harrison Eiteljorg, II, "Managing the Content of AutoCAD® Models with Layers," XXII, 2; September, 2009; csanet.org/newsletter/fall09/nlf0901.html. The approach adopted by Mr. Tucker and Mr. Wallrodt was designed to correct shortcomings those authors had found in the approach outlined in the CSA Layer-Naming Convention and the subsequent CSA Newsletter article. The authors specifically noted (1) the difficulty presented by the convention to users who were not very familiar with AutoCAD and (2) the assumption of the convention that layer names should reflect analytic conclusions. At the time of excavation, of course, those analytic conclusions will not have been reached.

At this point I must interrupt the text to suggest, as strongly as possible, that readers examine the CSA Layer-Naming Convention. The discussion that follows assumes an understanding of the system underlying the convention. That system descends from the old (supposedly originating in 1973) idea of the GREP search for UNIX files. GREP is an acronym that refers to the first of various systems to search text for specific characters and/or other characters that need not always be specified but can be adequately determined with a pattern description. That is, GREP searches permit text to be searched for patterns, not just specific words or parts of words, and that system underlies AutoCAD's layer-searching possibilities as well as the CSA Layer-Naming Convention.

Assuming you have reached this point with an understanding of the CSA Layer-Naming Convention, I want to explore the two criticisms of the convention that prompted Mr. Tucker and Mr. Wallrodt to seek an alternate. They argued that the convention was too complicated for most users of AutoCAD, users not accustomed to the complex interface and not wanting to be. They also argued that the convention assumed a state of analysis not available in the early stages of work.

I believe that Mr. Tucker and Mr. Wallrodt have tossed the proverbial baby out with the equally proverbial bathwater, and I wish to explain why in what follows.

First is the objection that the system assumes an analytic phase that must follow in time the excavation phase in which they were working. This text is from the convention document: "The layer-naming scheme described above is an analytic scheme and many of its specifications will not work well for excavations that are in progress. Too much of the final analytic process is presupposed by the system; in addition, it is unrelated to excavation processes, trench locations, excavation dates, etc. As a consequence, most excavators need a different system, one that separates material according to excavator, season, trench, and the like. Applying the same principle of using both the character and its position in the name, a system appropriate for any excavation can be worked out -- and documented."¹

The document continued with a series of examples of naming systems appropriate for the excavation stage of a project; the examples were intentionally rather general, since each excavation would have its own necessary data to be included in any layer-naming system. (The document also contained an explanation of a change in AutoCAD that permitted a layer-naming system that used, for any single layer, names appropriate both for the early, excavation phase and for the later, analytic phase.) Thus, I think the rejection of the convention on the basis of the argument about analysis phase versus excavation phase is specious.

The more valid argument, in my view, is that using the searching systems available in AutoCAD requires that a user be very familiar with the program and relatively comfortable using the software. It is not trivial to search for layers according to the system outlined in the convention. One must know exactly how the layer names have been constructed and what all the theoretical possibilities are. The system constructed for the PARP:PS project as an alternate was intended to make certain that, "group filters and layer names are clear and accessible to anyone who is expected to access the model directly."² [Emphasis added.]

The goal of clarity for all is certainly desirable. However, does what might be called simple clarity require that the hands of the expert user of AutoCAD be tied, preventing him/her from operating the program to its maximum utility? (Fully understood, the CSA system requires only this: that every meaningful distinction between or among layers be expressed with the same number of characters for each layer name so that the search system will function properly.) I think not. Indeed, that is why I think the baby went out with the bathwater.

If one examines the system created for PARP:PS, one finds a system of layer prefixes that conforms naturally to the CSA Layer-Naming Convention idea of permitting searches by both character and pattern. Each layer prefix consists of two letters followed by an underscore and a stratigraphic unit number (a five-character number). BUT some layers show features drawn by hand, while others show features drawn with total-station data. As a result, some layers have a p following the stratigraphic unit number to indicate that the features were drawn by hand; others have the characters ts following the stratigraphic unit number to indicate the use of a total station. (Other layers have no characters following the stratigraphic unit number to indicate the origin of the data; they are, if I have understood correctly, those layers for which only either hand-drawing or total-station data were used, but not both. Thus, a layer name beginning with SU_59006 only — no following p or ts — may have been drawn by hand or with total-station data while SU_59006p contains features explicitly drawn by hand and SU_59006ts has features explicitly drawn with the aid of a total station.) Were the convention used here, there would be no SU_59006 layer permitted; the CAD technician would be required to indicate whether the features were hand-drawn or drawn with the aid of a total station, something I would prefer in any case. In addition, the abbreviation for hand-drawn features would consist of two characters (perhaps hd in place of p) to make the number of characters the same for layers with hand-drawn and total-station data. That approach would permit a sophisticated user to search for (and find) layers by stratigraphic unit number OR the unit number plus the indicator for total-station data or hand-drawn data, or all units containing hand-drawn data, and so on. (Although it seems to me that the use of total-station data versus drawing by hand should be included, a layer without such a specification could have virtually any simple, two-letter designation that the data source is undefined — ud or zz, for instance. Such layers would easily fit the system.) The PARP:PS system, though, does not permit such searches to be effective. Thus, I would argue that the new system prevents an experienced user from taking advantage of AutoCAD's search possibilities for no good reason.

I must acknowledge here that the PARP:PS additions to the basic skeleton of layer names just discussed complicate the issue considerably. However, it seems to me that the creators of the system used in the PARP:PS project determined early on to go another way and did not look back. It does not seem to me that they made any attempt to try to provide both their goal of clear and accessible layer names and layer names that permit AutoCAD to be used to its fullest extent by those who know how to make it sing. Was it necessary to throw away the most effective use of layers in AutoCAD in order to make the names easy to understand? I really don't think so.

The system developed by Mr. Tucker and Mr. Wallrodt also uses layer groups (something relatively new to AutoCAD and therefore not discussed in the Layer-Naming Convention). Groups may be defined by rule (automatic assignment determined by layer names) or by explicitly assigning a layer to the proper group(s). That is, the AutoCAD user may assign layers to the group(s), or the user may build a rule that determines the group(s) to which any layer is assigned. Note that the rule can take advantage of the CSA Layer-Naming Convention. That is, had a system been developed that used the CSA Layer-Naming Convention, groups could have been defined so that all layers were placed in the correct group(s) automatically, without requiring any action on the part of CAD users. The PARP:PS system, on the other hand, requires that someone explicitly add layers to the appropriate group(s), making the groups more error-prone. This, too, suggests that a more careful approach to naming layers, one that used the underlying principles of the CSA Layer-Naming Convention, would have been more desirable for the PARP:PS project.

Since this is a matter that clearly does not lead to easy agreement, readers' responses and comments would be greatly appreciated.

-- Harrison Eiteljorg, II

Notes:

1. Eiteljorg, Harrison II, "CSA Layer-Naming Convention," (http://csanet.org/inftech/csalnc.html — section entitled "A Layer Naming System for Excavation Models," first paragraph. Accessed 6 September 2013.) Return to text.

1. Tucker, Gregory and Wallrodt, John, "Rethinking CAD Data Structures — Pompeii Archaeological Research Project: Porta Stabia," (http://csanet.org/newsletter/spring13/nls1302.html, paragraph 9. Accessed 6 September 2013.) Return to text.