* reconsider use of Thy_Info.get_theory: only works with batch-build, not within PIDE session;

  + consider in Build_Thydata:  Thy_Hierarchy.insert_errpatIDs

  + consider in Biegelinie.thy: KEStore_Elems.add_thes, 

* MW: check uses of Unsynchronized.ref vs. Synchronized.var;

* MW: clarify/eliminate Isabelle/Scala add-ons (presently unused)

    diff -r /home/makarius/isabelle/repos-Isabelle2021/src/Pure/build-jars ./src/Pure/build-jars

    11a12

    >   src/Tools/isac/BridgeJEdit/isac.scala

    diff -r /home/makarius/isabelle/repos-Isabelle2021/src/Pure/Tools/scala_project.scala ./src/Pure/Tools/scala_project.scala

    76a77

    >       "src/Tools/isac/etc" -> Path.explode("isabelle.isac"),

* Clarify symmetric rule: Thm.apply_attribute Calculation.symmetric thm context (!?);

* Is it possible to eliminate union_overwrite, in favour of more standard data

  add/merge discipline?

* What is the purpose of "#numeral" instead of plain numeral?

* Check/clarify Context.theory_name vs. Context.theory_long_name.

* Eliminate mutable Rewrite_Ord.rew_ord' (!?);

* WN: eliminate ThmC.numerals_to_Free, use existing Isabelle/HOL representation

    - clarify role of type "real" vs. "float" (see theory "HOL-Library.Float");

* WN: clarify Rule.Thm ("minus_divide_left", ...): Should this be @{rule_thm_sym} antiquotation?

* WN: check remaining MethodC.prep_input that use a different theory (e.g. "Diff"):

  Is this really required, or can we just use the 'method' command?

* WN: remove always empty arguments in MethodC.prep_input and Problem.prep_input

  (following the "TODO" comment);

* WN: trim-down MethodC.prep_input and Problem.prep_input to what is really needed.

* WN: eliminate global flags like "trace_on", replace Unsynchronized.ref by

    ML \<open>val rewrite_trace = Attrib.setup_config_bool \<^binding>\<open>rewrite_trace\<close> (K false);\<close>

* WN: Avoid Thm.get_name_hint --- somewhat fragile.

* WN: eliminate ThyC.to_ctxt, use Proof_Context.init_global inline to make more

  clear where the normal Isabelle context-discipline is bypassed;

* WN: more direct logical foundations wrt. Isabelle/HOL, eliminate many axiomatizations

    - quite often "axiomatization ..." can be turned into "lemma ... by auto"

      without further ado;

    - sometimes this requires to use more specific types / type classes;

    - sometimes this requires to use proper definitional mechanisms (e.g. 'primrec', 'fun');

    - a few "hard" cases will remain, to be reconsidered eventually (e.g. differentiation);

* WN: "fun pr_ord" is not required if used with @{make_string}, @{print}, @{print tracing};