[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: snmpconf Snapshot draft

Thanks for describing how the overloaded variables are sorted out.

PS: I know it is common, and we should probably use this overloading, but I 
am doubtful as to how much it is worth.

At 03:11 PM 2/26/01 -0800, you wrote:

>Much has been done in the last draft. I've included a list of notable
>changes at the end of this message.
>The change most worthy of some attention is the addition of
>loosely-typed variables to the language. Loosely-typed variables
>are an attribute of every interpreted language I've been examining
>(Perl, TCL, Javascript, Python). With this we get automatic
>conversion as appropriate from integers to strings and back, no
>need to make size decisions and cast between different sizes of the
>same type, simpler accessor function APIs (also fewer functions
>needed), no need to declare type, etc. All of these culminate in the
>real goal: shorter, easier to write scripts.
>For example, a popular idiom for us is to construct OIDs. So what once
>     char oid[STRINGLEN];
>     sprintf(oid, "ifAdminStatus.%d", index);
>     writeVar(pdu, 0, oid, INTEGER, 2, 4);
>     writeVar(pdu, 0, "ifAdminStatus." + index, INTEGER, 2);
>What this all looks like is:
>   var x, n = 5, str = "IETF";
>       (declaration doesn't include type)
>These are OK:
>   x = "Hello World";
>   x = 20;
>Note the following:
>"String" is of type string
>55 is of type integer
>"55" is of type string
>Integer("55") is of type integer (the resulting value is 55)
>If you perform A + B and either A or B is a string, it converts both
>to strings and appends them. Hence:
>    var index = 5;
>    puts("ifAdminStatus." + index); ==> ifAdminStatus.5
>    /* puts is just for demonstration. It's not in our library */
>If both are integers, it adds them:
>    var index = 5, num = 10;
>    puts(index + num); ==> 15
>All other arithmetic operators convert operands to integers, so:
>     var str = "10", num = 5;
>     puts(str * num); ==> 50
>There's a detailed description of the conversion rules in the
>description of "the var class".
>The other changes include the following from Jon's todo list:
>3, 5, 7, 9, 10, 15, 16, 17, 18, 19, 23, 24, 25, 28, 29, 35, 36,
>37, 38, 39, 40, 48.
>Here's a description of the changes:
>- Added policyParameters and getParameters()
>- Error to use a varbind that wasn't set
>- Can we get a reference for the biblio for EBNF?
>- Mention:
>     implmentation specific limits on scratchpad variables per policy
>     implmentation specific limits on global scratchpad may be lower
>- Consider (in examples), having descriptor-style OIDs not in quotes so
>   they aren't confused
>- Be very explicit that examples are using descripters despite the fact
>   that this is not the "over the wire" representation.
>- More text about "this element" being index, context + address.
>- policyGroup changes
>     New "all at once" semantics
>     When b trumps a, b runs for a while, b now terminates, we must
>       re-execute a IMMEDIATELY
>     Make things happen immediately when precedence changes
>- incorporate (by value) schedule mib with changes
>- Need to define element "0.0" for "this system" (think of it as
>   policies for scalar elements
>- Make changes for pass by reference
>- Change to C++ reference
>- Add automatic initialization
>- Add newPDU()
>- Explicit element addresses as varargs for roleMatch and capMatch
>   for example, when processing DLCI, what is role of containing port?
>- Add signalError() accessor function
>     and enumeration per execution context (also for abnormalTermination)
>     and pmSignalErrors gauge
>- Define terms like execution context and "this element"
>- Added loosely-typed variables
>- Sscanf
>- replace string "access functions" -> "accessor functions"
>- Policy state object for showing inactive due to precedence.
>- Security issue with deferring the loading of a policy based on
>   lack of a role. Policy may not be able to be loaded in time to
>   prevent havoc. Added in mib and in security considerations.
>- PolicyAdminStatus
>- iv() and ic()
>- Integer() and String()
>- chr() and ord()
>- Delete strncpy(), strncat() and atoi().
>- string substr(string s, int start, int len);
>- string replace(string s1, int start, int len, string s2);
>- reordered and renamed arguments to oidsplice
>- reordered arguments in readVar/writeVar
>- Fixed a few bugs in grammar for lexical elements and added octal
>   encodings and escape sequences.
>- Deleted getint() and setint() as unnecessary (getvar() is now just
>   as simple)
>- Updated declarations of many accessor functions
>- Other typos, editorial changes, formatting, etc.