RFD: Misleading log --graph online

Consider these snippets of output: git log --graph --oneline --decorate --all

* ca19d2e (upstream/trunk, upstream/master, svn/trunk) update upload data
* cd2cb19 update download url
...
* ab11781 new filelocator module
* 8125929 fix: dodata->doplot
| * e234268 (svn-tags/pyx_0_11) tagging release 0.11
| * 580cbdf (svn-tags/pyx_0_11@3128) documentation update...
...
| * ffa2e8f *** empty log message ***
| * 90dd12c *** empty log message ***
| * 2c29cab (svn-tags/pyx_0_10) tagging 0.10
|/
* 42fdff5 (tag: svn/pyx_0_10) some updates to the webpage
* fac723b some adjustments to keep support for a broad range...
...

Obviously, there is a side branch svn-tags/pyx_0_11 which forked off 42fdff5, right?

No, in fact 90dd12c is the root commit of that branch! 2c29cab is the tip of svn-tags/pyx_0_10.

In other words, our graph output can be very misleading in a case like that. I would suggest to mark a root commit with "o" like we do for boundary commits. (tig uses "o" or "*" for ordinary commits and "I" for a root.)

But I'm not sure that we don't create misleading output like that in other situations also, like for A@{1}..A where we produce "virtual roots" as well.

The main problem appears to be that "*" does not convey any connectivity information. Wondering whether we could use graphic symbols from some charset...

Michael