[146890] in North American Network Operators' Group
Re: Network device command line interfaces
daemon@ATHENA.MIT.EDU (Keegan Holley)
Thu Nov 24 23:52:02 2011
In-Reply-To: <EA43F3C127260242916718667C11BAAA03B25C@WINEXP12.internal.local>
From: Keegan Holley <keegan.holley@sungard.com>
Date: Thu, 24 Nov 2011 23:51:05 -0500
To: Jonathon Exley <Jonathon.Exley@kordia.co.nz>
Cc: "nanog@nanog.org" <nanog@nanog.org>
Errors-To: nanog-bounces+nanog.discuss=bloom-picayune.mit.edu@nanog.org
That's kinda what I was talking about. That command isn't that bad actually=
. MQC and juniper firewall filters (in set mode) are no simpler. The annoyi=
ng part is the obscurity.=20
Sent from my iPhone
On Nov 24, 2011, at 11:38 PM, Jonathon Exley <Jonathon.Exley@kordia.co.nz> w=
rote:
> Yeah, I guess Cisco IOS isn't that good an example of a consistent syntax.=
Others do it better - Junos sets the ASN with the 'routing-options autonomo=
us-system' command, and TiMOS uses 'router autonomous-system'
>=20
> My rant wasn't about having to deal with new CLIs but about the lack of CL=
Is in those devices that seem to prefer menu based UIs (text or web), and CL=
Is that have nasty commands. Check this out:
>=20
> add flow fid-5-5 EVC-30600-Data codefault enable multi swap 99968000 10003=
2000 1024 1024 5000 ctag push 15-0 stag none
>=20
> Now what does that string of numbers mean? It's the Adva 825 way of specif=
ying the CIR and EIR for a flow but I can never remember what each position r=
epresents.
>=20
> Compare this to TiMOS:
>=20
>=20
> sap-ingress 93 create
>=20
> description "Test LNS"
>=20
> queue 1 create
>=20
> rate 2000
>=20
> mbs 25 kilobytes
>=20
> exit
>=20
> This creates a queue with max rate 2000 kbit/s and a max burst size of 25 k=
B. It's much easier to read than the Adva config, because each parameter is l=
abelled.
>=20
> The Adva CLI isn't actually all that bad, but it's possible that had their=
developers had some sort of usability guide when they wrote the OS then the=
y might have done things better.
>=20
> I was hoping that there was already some sort of usability guide around th=
at could be shown to the manufacturers with a "please read this" note attach=
ed. Is anyone aware of such a thing?
>=20
>=20
> Jonathon.
>=20
>=20
> From: Keegan Holley [mailto:keegan.holley@sungard.com]
> Sent: Friday, 25 November 2011 4:12 p.m.
> To: Jonathon Exley
> Cc: nanog@nanog.org
> Subject: Re: Network device command line interfaces
>=20
> I may have a different opinion here, but I not sure I'd call any CLI easy t=
o work with. Cisco's training machine is so efficient that some learn IOS b=
efore leaving high school, so the fact that we all consider IOS easy to work=
with is relative. Just look at the "router" command. Most of us know that=
this is cisco's way of enabling protocols, but I would hardly call this int=
uitive if I didn't know it already. Then it's different for each protocol. S=
o "router BGP #" starts the BGP process and sets your local AS number (very i=
mportant). "router eigrp #" starts eigrp and sets a different AS number that=
doesn't really count (also important). "router ospf #" just sets a process I=
D in case you want to run multiple instances. There's also a config mode au=
tonomous-system command but that only counts if your running EGP which is st=
ill in the CLI but isn't supported and doesn't start. Then there's all the d=
ifferent things you can/must do with access-lists because they were too lazy=
to code a different sort of filter. Remember CBAC? Did I mention this is t=
he CLI we like? I don't mind wrestling with a new CLI because it's all rela=
tive. Most have read at least one cisco book and probably one juniper book s=
o those CLI's are considered standard and all their sins are forgiven. Most=
of us have not gone through, training with extreme, enterasys, 3COM, netgea=
r, foundry, fortigate, etc. etc. etc. So those become the PITA CLI's and su=
ddenly non-standard commands and bad help menus become a crime again. I do f=
ind text-based menus obnoxious, unless it's a linux box and the text menu is=
a curses interface. In that case it's super-cool and I'm even willing to p=
lay games with text based menus.
>=20
> 2011/11/23 Jonathon Exley <Jonathon.Exley@kordia.co.nz<mailto:Jonathon.Exl=
ey@kordia.co.nz>>
> Does anyone else despair at the CLIs produced by networking vendors?
> Real routers use a CLI that is command based, like IOS, TiMOS or Junos. Th=
ese interfaces work well over low bandwidth connections (unlike web interfac=
es), can work with config backup systems like RANCID, have a (mostly) consis=
tent structure and good show commands.
> However vendors of low cost routers/switches/muxes seem to take a stab in t=
he dark and produce some really nasty stuff. I have a personal hate of text b=
ased menus and binary config backup files.
> Doe this p*** off anyone else? The business part of the company says "This=
device is great! It's cheap and does everything." However the poor sap who i=
s given the task to make it work has to wrestle with a badly designed user i=
nterface and illogical syntax.
> Maybe the vendors need some sort of best practices guide for what manageab=
ility features their kit needs to support to make them acceptable to the mar=
ket. Does anyone know if there is anything along these lines?
>=20
>=20
> Jonathon.
>=20
>=20
> This email and attachments: are confidential; may be protected by privileg=
e and copyright; if received in error may not be used, copied, or kept; are n=
ot guaranteed to be virus-free; may not express the views of Kordia(R); do n=
ot designate an information system; and do not give rise to any liability fo=
r Kordia(R).
>=20
> This email and attachments: are confidential; may be protected by privileg=
e and copyright; if received in error may not be used, copied, or kept; are n=
ot guaranteed to be virus-free; may not express the views of Kordia(R); do n=
ot designate an information system; and do not give rise to any liability fo=
r Kordia(R).
>=20
