From: arne@vajhoej.dk   
      
   On 12/7/2025 8:17 AM, Craig A. Berry wrote:   
   > On 12/6/25 5:46 PM, Arne Vajhøj wrote:   
   >> On 12/6/2025 4:28 PM, Craig A. Berry wrote:   
   >>>                                  
    Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â Â    Even if user-   
   >>> written lexicals were possible, you couldn't really use them to create   
   >>> or manage very interesting data structures given that DCL symbol values   
   >>> are limited to 1024 characters.   
   >>   
   >> I believe it is 8192 today.   
   >   
   > Current online help for LIB$SET_SYMBOL says 1024. Current RTL manual   
   > says 4096.  Dunno what's right.   
      
   $ HELP SYMB   
      
   says:   
      
      
              The value of the symbol can be a 32-bit integer or a string   
         with 1 to 255 characters (for VAX) or 1 to 8192 characters (for   
         Alpha and I64 systems).   
      
      
   And more important DCL "confirms":   
      
   $ typ maxsym.com   
   $ write sys$output f$getsyi("version") + " " + f$getsyi("arch_name")   
   $ s = ""   
   $ i = 0   
   $ loop:   
   $    if i .ge. 8100 then goto endloop   
   $    s = s + "X"   
   $    i = i + 1   
   $    goto loop   
   $ endloop:   
   $ write sys$output f$length(s)   
   $ exit   
      
   $ @maxsym   
   V9.2-3   x86_64   
   8100   
      
   $ @maxsym   
   V8.4-2L2 Alpha   
   8100   
      
   Well - the above can actually only do 8163 not 8192. Over 8163   
   it gives:   
      
   %DCL-W-BUFOVF, command buffer overflow - shorten expression or command line   
      
   but I buy the 8192 for the pure symbol.   
      
   >                                 But with people using scripting   
   > languages to process massive vectors to train LLM models, it all seems   
   > pretty puny.   
   It is.   
      
   Arne   
      
   --- SoupGate-Win32 v1.05   
    * Origin: you cannot sedate... all the things you hate (1:229/2)