For conversations, I recommend that we use short label/terminal Y SNP names, but within the R haplogroup that we don't use the single letter "R", but rather use R1a, R1b or R2 as is appropriate. These haplogroups are only distantly related and both R1a and R1b are quite large.
Some may ask, why three letters instead of four, etc. ? I think the answer is that R1a and R1b are de facto standards both across the academic and hobbyist community. Most people know what they are. I am against using long phylogenetic labels in conversations.
As a case in point. This new paper was just published last week:
"A calibrated human Y-chromosomal phylogeny based on resequencing" by Wei, et al.
http://genome.cshlp.org/content/early/2012/10/04/gr.143198.112
If you open the free PDF text file and scroll down to the last page you will see very clearly that R1a, R1b and R2 are clearly broken and levels below that are just referred to as son, father, etc. No standard is shown below the R1a and R1b depictions.
Try this exercise. Imagine you are a novice and you want to do research. Hopefully, you will do some internet searches. Try googling R1b. Now try googling R. Which is more helpful?
R1a, R1b, R2 are commonly accepted terms that have meaning. R1a and R1b are too big to clump together.
Now, let me switch gears. I am NOT for discontinuing long phylogenetically intelligent haplogroup labels altogether. Apparently FTDNA will do this. However, I hope they, or at least ISOGG will still maintain a phylogenetically intelligent haplogroup labeling system in addition to short terminal SNP labeling.
Y DNA trees require some intelligent recordkeeping, or more specifically data design. Hopefully, we are beyond always having to have a bit based only graphic picture of the Y DNA tree. Every branch on the tree needs a tag that tells you what limb it grew out of. Every branch also needs data tags for multiple children branches or twigs. This is phylogenetic intelligence. The long haplogroup labels provide this and therefore are very useful in data analysis, statistical research and the like.
For instance, you can select "R1b1a2" with a wild card character and get all versions of people who are "R1b1a2" or deeper, regardless of the level of SNP testing they've actually gone through. Ysearch allows you to do this today, or at least the last time I tried this.
What about R-L176.1 and R-L176.2? They would sort out right next to each other in a report. This is a bit misleading, particularly to a novice. R-L176.1 is in R1a and R-L176.2 is in R1b and so these guys aren't closely related at all and probably 20k years apart.
I'm going to start following a conversational nomenclature the U106 project admin recommends. For myself, I'm going list that I'm R1b-L21 in postings. Actually, although not many will know what it is, I'm going to list myself as R1b-L21>L705.2 since L705.2 is my true terminal SNP and most people will get a pretty good idea where I fit by including L21.
I don't have answer for this problem, except to demonstrate why long phylogenetically intelligent labels are still important.
The problem is that all of these terminal haplogroup labels - R1b-M529, R1b-S145, R1b-L21, R1b-Z245, R1b-L459, R1bg-rs11799226... are the same. Did you know that? Z245 and L459 are phylogenetically equivalent and the rest are just different names for the same SNP. I wonder what FTDNA will do for a person who is Z245+ but didn't test for L21 specifically. There is no need for them to test for L21 but I guess they will be a different haplogroup.
Here is the lesson learned.
Just like pots are not people, SNPs are not clades. SNPs are just tags or signposts on branches of they Y DNA tree of paternal lineages. We don't know if an SNP is part way up a large branch, or at the base, or what. This will be dynamic that is always changes as more is discovered. In a sense, the long phylogenetic labels are more appropriate. They describe the branch, regardless of the number of SNPs on it. STRs can effectively define a branch as well. A case in point would be 490=10 among R1b-SRY2627 (M167) people.
The net is we need both. Long phylogenetically intelligent labels for analysis, as well as short acronyms for ordinary conversation. Even if the formatting is different the long phylogenetic names are still important so a display/charting program can draw the tree.
This is exactly what GEDCOM is. Perhaps we should store the SNPs in a GEDCOM file to drive the tree charts. As we need precise, hiearchical analysis, we can use the GEDCOM coordinate definitions rather than the multi-letter/digit approach. This wasn't idea but I think it is worth considering.