Within the Parcel Fabric, ESRI's solution for parcel maintenance, there are two distinct methods with which you can handle dimensions: labels or feature linked annotation. Each method has its benefits and limitations, as well as defenders and detractors.
Panda Consulting has historically leaned towards the labeling side when using simple feature classes for your data model. However, the recent change in how we share our parcel data has us thinking. "Should we recommend Labels or Annotation?" recognizing there is no perfect solution.
When the primary client used for viewing GIS parcel data consisted of ArcGIS for Desktop or ArcReader, it was easy to assume that the software could read and create labels using the core labeling engine technology. However, since the new paradigm for data publication is that any device, including smart phones and tablets, will be used to display the data, that assumption that the client has the local processing power or software to generate these labels forces us to rethink this assumption.
Storage of Dimensions in the Parcel Fabric
Whether you decide to use labels or feature linked annotation, all dimensional information is stored in the parcel fabric within the parcel lines feature class (similar to COGO attributes on simple feature classes), where it defines the core geometry of the individual polygon. When the attribute is modified, the corresponding label or feature linked annotation is also modified.
Publication of Dimensions
If we were to keep the data within the Parcel Fabric and never share the dimensions with others, we would not need to decide between the two. However, we now use an Extract, Transform and Load (ETL) process to take the data from the Parcel Fabric, a complex structure, into simpler lines and polygons feature classes for publication. Since the creation of feature linked annotation requires an extra process over just storing the data in the line features and using labels. Storage and Maintenance = Advantage Labels
Feature Linked Annotation
Annotation actually has a small advantage in that, since feature linked annotation is not actually stored in the Parcel Fabric but in an annotation feature class that is related to the parcel lines, the ETL process is as simple as copying the annotation feature class into the publication database. The downside to annotation has always been that it is created for a single scale, whereas labels can be set to change size and display properties as the scale changes. If multi-scale annotation is desired, it must be created many times. In addition, annotation takes up disk space to store.
The use of labels is complicated by the manner in which the Parcel Fabric stores those features used to produce the dimensions. In the Parcel Fabric, all features having the same geometry type (points, lines and polygons) are aggregated into a single geometric feature class. In addition, since lines and polygons are maintained as being either active or historic, only lines that are directly related to "active" tax parcels are used for labeling purposes. Therefore, during the ETL process, a selection query must be used to only extract those lines that define active tax parcels.
That means, you must maintain the "Type" column in the Parcel Fabric's Parcel Lines feature class to remain correct as you modify the parcel type in the polygon feature class. ArcGIS 10.2.1 correctly sets the type attribute field during the parcel creation process, but, if you change the parcel type after creation,you must manually change the corresponding lines for the data to be correct. While this is not difficult to do, it would be nice if ESRI would fix this problem and make the change automatic. Extraction of Dimensions = Advantage Feature Linked Annotation
Consumption of Dimensions on the Web
Once you are finished with your maintenance and export your data for publication, there is still the issue of how the clients will consume the data. Unfortunately, ArcGIS Online does not currently support annotation features as standard feature services so, the most efficient way to currently serve out dimensions to your web maps is by creating a cached map service that sets display scales automatically for the the annotation or labels. Creation of the tiles for this cached map services is required for both feature linked annotation and labeling. Publishing and sharing on the Web = Tie
What does this mean?
So, what does this all mean? When planning your transition to the Parcel Fabric, be sure to think about more than what you have previously decided, be sure to think about how you will be serving out your data. Understand that, no matter which method of storing dimensions, you will still probably be creating a cached web service and serving out the cache rather than the actual features.