sysmoISIM-SJA2 clock stop discrepancy

[Mychaela Falconia]

Hello esteemed masters of Osmocom and Sysmocom,

As I am doing some experiments with sysmoISIM-SJA2 cards in order to
make a contingency plan in case I am not able to buy cards from Grcard
beyond the 5 sample cards I got in February (my contact person there
hasn't been responding to email for the past 3 days), I have noticed a
discrepancy between this card's ATR and the file characteristics byte
it returns in response to SELECT in the GSM 11.11 SIM protocol,
regarding allowed levels for clock stop.

My sysmoISIM-SJA2 cards (bought from the webshop in late Jan into
early Feb) return the following ATR:

3B 9F 96 80 1F 87 80 31 E0 73 FE 21 1B 67 4A 4C 75 30 34 05 4B A9

The TA byte for T=15 indicates allowed voltage classes and allowed
levels for clock stop.  On this card this byte equals 0x87, meaning
all 3 voltage classes (good), but also indicating that when the clock
is stopped, it should be HIGH.

However, the file characteristics byte returned in response to SELECT
(both MF and DF_GSM) in the GSM 11.11 SIM protocol is 0xB1; decoding
this byte per 3GPP TS 51.011, it says "clock stop allowed, no preferred
level".  Thus the file characteristics byte says that there is no
preferred level, yet ATR says that HIGH level is preferred - or not
just preferred, but required?  (7816-3 seems to say it is only a
preference, not a requirement; I seem to recall some other spec that
said it's a requirement, but I can't find it now.)

This discrepancy poses a potential problem if these cards are to be
used in classic GSM/2G dumbphones whose original firmwares are based
on the reference from TI.  That reference fw code does not look at ATR
for the purpose of determining allowed clock stop levels (most classic
SIMs didn't have that T=15 TA byte in their ATR, it became commonly
present when UICC/USIM stuff came along), instead it looks at the file
characteristics byte returned in response to SELECT of DF_GSM.
Furthermore, if the file characteristics byte says that there is no
preferred level for clock stop, TI's reference fw configures the hw to
leave the clock LOW during idle.  I can only reason that other classic
dumbphone firmwares (from other chip vendors) may very likely do the
same: GSM SIM specs (11.11 and 51.011) don't require ME implementations
to look at T=15 bytes in ATR, instead they direct MEs to follow the
file characteristics byte.

Can someone from Sysmocom officially confirm whether or not it is OK
to operate sysmoISIM-SJA2 cards with clock stop at LOW level, contrary
to ATR asking it to be HIGH?  I have done a limited test of putting
one of these cards into an FCDEV3B (with this aspect of the firmware
left unmodified, so the clock line was low during idle) and the SIM
interface appeared to still be alive after some deep sleep (clock stop)
cycles.  However, it was a very limited test, and I don't have my own
network set up yet to make a more thorough test - and in any case, an
official confirmation would be much better than anecdotal observations.

TIA,
Mychaela